In the age of programmatic media, marketers are beginning to realize the promises of audience-based marketing. At the same time, the modern consumer's meandering purchase path across multiple touchpoints on multiple devices has made one-to-one communication even harder. One solution to both marketing and measurement across touchpoints is cross-device tracking.
When it comes to understanding cross-device tracking, you're likely familiar with the two primary methods: deterministic and probabalistic matching. (If not, here's the difference: deterministic relies on some specific piece of known user data, say a Facebook log-in or an email address, across devices, while probabalistic matching approximates when a user is present on multiple devices based on usage patterns.) More and more, companies are blending the approaches, often using some deterministic data to train their probabalistic matching algorithms.
In sum, it's getting complicated. While you may feel comfortable with the difference between these approaches, you may not have the same clarity around the various methods and how close you are to reaching the ultimate goal of one-to-one communication.
Understanding Probabalistic Matching: Accuracy vs. Precision
With deterministic matching, your main question is scale: you can assume your matches are accurate, but you can only account for so many log-ins. Your primary metric when comparing providers should probably be reach.
When it comes to probabalistic matching, reach is an important metric, but it isn't the only important one. You also want to understand how effectively the provider determines their matches. Most providers measure two things: accuracy and precision. Accuracy and precision are simple enough terms, but in this context they have very specific meanings that aren't immediately apparent.
Accuracy in probabalistic matching refers to the number of correctly identified matches AND the number of correctly identified non-matches. If you've done any shopping around for a cross-device tracking company, you may have wondered why they can boast such high accuracy numbers. It's because accuracy includes both the matches they have found and the matches they have definitively identified as inaccurate. A high accuracy number does not indicate a large reach.
Precision, however, is what you probably thought they meant by "accuracy." Precision refers to the proportion of matches confirmed to be correct out of all predicted matches. For most marketers, this is probably the more important metric and the one you should focus on.
Precision and reach should your two big concerns. Is there anyone in the industry who can offer both? Yes; and it's exactly who you think it is.
Accuracy Meets Scale: Within the Walled Gardens
It's only natural that the biggest players in ad tech (and digital identity) would become the big players in cross-device tracking.
When Facebook relaunched Atlas in late 2014, it was with the promise of a new future of cross-device advertising that did not rely on cookies, but rather on the log-in data of Facebook's extensive user base. Facebook's Atlas supports both offline attribution measurement (a major challenge for many marketers), with a product they call "Offline Actions," and "Path to Conversion," a tool for measuring a consumer's multi-touch path to purchase.
In the summer of 2015, Google announced their answer to Atlas, a cross-device measurement solution that, unlike Facebook, uses probabilistic modelling. Deterministic data is used in a limited way to inform the matching algorithms. It's also important to note that Google does not support cross-device tracking for advertising, only for measurement. If you want to understand the customer's journey after the fact, this tool can work, but if your concerns include frequency capping across devices or targeting, this isn't the solution for you.
The Future Is Bright
Ultimately, there's still a long way to go when it comes to accurately measuring the customer journey across the many devices that people now use on a daily basis. Understanding the tech in play (and its limitations) is crucial for marketers looking to improve their audience-based audience.
