• by Tyler Loechner , Staff Writer @mp_tyler, December 3, 2014

The ability to run cross-device campaigns has become a staple promise of programmatic ad tech providers, but matching the devices in a privacy-friendly manner -- i.e. knowing that Smartphone A user is indeed the same as Desktop A user without using personally identifiable information -- has been cross-device’s “yeah but” roadblock. 

“In today’s always-on, always-engaged world, lack of predictable cross-device connections can be a serious barrier to success for marketers and content providers,” wrote Tapad, a cross-device marketing tech provider, in a recent blog post.  In short, if marketers are going to spend money to follow audiences around, they want to make sure they are following the right people.

When it comes to cross-device advertising, there are two approaches: probabilistic device pairing and deterministic device pairing. Daryl McNutt, former VP of marketing at Drawbridge and current CMO Adaptive Media, wrote a piece for Real-Time Daily in April explaining the probabilistic and deterministic approaches.

The deterministic method uses personally identifiable information -- sometimes called PII -- from sites that require a login on all devices, such as Facebook. McNutt wrote that deterministic cross-device pairing leads to successful matching of at least 95%. However, “privacy concerns are by far the largest drawbacks to this approach,” wrote McNutt.

Probabilistic modeling uses no personally identifiable information, so the privacy concerns are much less prevalent. However, since there is no PII to “prove” that the consumer is the same across devices, the success rate in matching is lower.

But Tapad, a cross-device marketing tech provider, this week made an announcement that may change the way marketers view probabilistic modeling.

Nielsen measured the accuracy of a sample of Tapad’s data set and concluded that Tapad successfully matches devices 91.2% of the time. That’s only slightly below deterministic matching success, per McNutt.

It must be noted that Tapad commissioned Nielsen to measure their accuracy, though they claim in a blog post announcing the results that it was an “independent assessment.”

Tapad has been in business since 2010, and its possible that their matching success only recently broke the 90% barrier. McNutt explained in his post from April that the algorithms from companies such as Tapad “recognize patterns and make predictions that become stronger over time.” 

Tapad’s announcement is noteworthy because it gives credence to “algorithms that learn.” It also suggests that the proverbial answer is in the data, and doesn’t always have to be served on a silver platter (i.e. logins).