The opportunity and challenge of mobile
Similar to cookies in fixed web environments, mobile device IDs can be used as a proxy to user IDs to create a consumer profile and engage with the consumer. Mobile device IDs are proving to be more reliable and effective than fixed web cookies in building a consumer profile and engaging with the consumer.
Although mobile device IDs are more reliable than cookies, mobile consumption is often driven by consumers seeking to achieve a specific episodic need. Mobile devices are not yet the primary mode of digital content consumption for the majority of consumers; so relying on mobile data alone would unduly limit the scope of data flowing into anonymous audience profiles. The type of actions consumer pursue on mobile also differ from those they pursue on fixed web applications. For example, a user may extensively research a product on mobile, but actually convert through the fixed web. The solution? Matching cross-device anonymous profiles and data, allowing us to pull data from online cookies for targeting on mobile and vice versa.
Connected-device user identification
With today’s new sources of information, the challenge is discovering and targeting relevant audiences from disconnected devices. Efforts are now focused on comparing how audiences are consuming content across devices, taking into account all of the different non-PII signals: IP addresses, locations, timeframe and time stamps, device make and model, device OS, screen size, affinities, interests, demographics, frequency and much more.
The ultimate goal is to use these different signals to create a device graph that identifies links between various devices used by the same consumer. Once these links are identified, digital stakeholders can port data from one device to another connected device, creating a unified anonymous consumer profile that can be used to create effective cross-channel marketing campaigns. This enables a wide variety of targeting and analytics, including cross-channel retargeting, sequential messaging, and the ability to measure the effectiveness of one channel over another.
Nuances of cross-device user identification
Using a device graph technology for cross-device user identification involves both art and science. While the device graph algorithm itself is a big step in the right direction, its success will depend on the richness of the data set that is fed into the graph.
The ability to identify relationships between devices with a high degree of accuracy based on probabilistic matches depends on the amount of data one has across all these devices. For companies that see billions of cookies and devices every month, the scale of information on every device in terms of behaviors, profiles and locations will enable these technologies to build powerful cross-device identification.
When dealing with device graph technologies, digital stakeholders need to keep three key metrics in mind. In addition to the match rates a device graph can generate, these metrics help understand the effectiveness of any device graph technology:
1. Accuracy: What is the accuracy of cross-device user identification when the graph is applied to a known data set of cookies and device IDs that are tied to the same consumer? In other words, what fraction of matches the algorithm declares correspond to devices belonging to the same user? Accuracy can only be measured with gold-standard testsets and posthoc analysis.
2. Confidence: What is the confidence associated with the models used to create the device graph technology? Confidence is a metric produced by the model and can be a proxy for accuracy. It is immediately available on each declaration and doesn’t require further analysis. It is simply how well the algorithm believes it is doing.
3. Reach: Over many examples, how many times can the device graph link two devices together with the same level of confidence/accuracy? This has implications for the usefulness of the entire cross-device enterprise.
Confidence, Accuracy and Reach
Digital stakeholders need to understand a unified anonymous profile of consumers across connected devices. The industry is making great strides in deploying unique technologies like the device graph to help facilitate cross-device consumer engagement. For these efforts to be effective, the technologies should be accurate, efficient, scalable and privacy-compliant. It is important to keep the combination of confidence, accuracy and reach in mind to create an effective device graph. All three go hand in hand and cannot be used in a silo.
High match rates derived using device graph technologies with low precision, without a high degree of confidence and without a high level of accuracy are as good as saying, “I am 100% confident the earth is flat!”
What is required by digital data owners is a well-constructed set of rules that results in a balance between accuracy and reach, and allows campaigns to be run effectively and at scale.