Google Adds Cross-Device Metrics To DoubleClick, Partially Answers Facebook’s ‘People’ Power

mohan nealEveryone knows Google needs to counter Facebook’s people-based marketing approach, unveiled with much fanfare last year as an answer to the twin measurement problems of device fragmentation and the crumbling cookie. On Wednesday it did so to an extent, making cross-device measurement of media exposures and conversions available in all DoubleClick ad products.

The announcement, expected Wednesday at DoubleClick’s Leadership Summit in Key Biscayne, Fla., will be accompanied by a flurry of other product news, including support for native ad serving and a programmatic guaranteed offering. Keep reading for details on all three initiatives.

Google’s cross-device solution is both more and less than Facebook’s. On the plus side for Google, it has more access to online conversion data through its search ads business. But the volume, accuracy and potential use cases of its cross-device data may be somewhat less than those of its rival.

Quality is a huge question with any cross-device measurement solution, and while Google has fewer persistently logged-in cross-device users than Facebook, it still has a lot (as many as 600 million, AdExchanger has estimated). That gives the company a huge base of so-called deterministic data, which is more accurate than probabilistic data, where the leading vendors such as Drawbridge and Tapad claim anywhere from 70 to 90% accuracy in matching devices.

But because it’s Google, privacy is bound to be an issue. To help address any concerns about its use of logged-in audience data, Google says it will only use its logged-in data to kick-start a probabilistic model, unlike Facebook, which is deterministic through and through.

“We use people who have signed in to Google accounts on various devices as seed data and we extrapolate from there,” said Neal Mohan, Google’s VP of video and display advertising.

cookoffAdditionally, Google’s is a measurement-only solution (i.e., cross-device targeting is not supported), and all data is displayed to the marketer in aggregate.

Google also offered a few case studies of marketers using the cross-device tracking capability. Mobile display ads drove 23% higher conversions for smart-home product maker Vivint when it used AdWords vs. single-device measurement. Fashion brand Charlotte Russe discovered it had 8.5% more conversions using Google Display Network (GDN). Also in GDN, La-Z-Boy tracked 31% more conversions.

Mohan said Google’s cross-device measurement solution has been in testing for about 18 months.

Native Ad-Serving Support

As AdExchanger previously reported, Google has begun supporting native ad serving in DFP in a move that could ramp up liquidity in the burgeoning native space.

“The idea is let’s make native advertising work in a way similar to how display works,” Mohan said. “The issue is, by definition native is different for every different publisher and every device.”

To address the fragmented standards in native and reduce the production burden on publishers and advertisers, DoubleClick says it can automatically generate, serve and report on the ads based on any IAB standard display ad format. For publishers that lack the resources or do not care to develop custom ad formats, DoubleClick will offer standardized native templates.

Direct Deals

As more fixed-price digital media is transacted programmatically, more companies are throwing R&D at direct sales tools. Google is no exception. It says programmatic direct transactions have doubled in the past year, and eight of its 25 largest publishers sell at least 10% of their impressions this way.

“Programmatic direct is incredibly important to global brands and large publishers,” Mohan said. “It’s the ability for publishers to set fixed prices for a chunk of inventory and ability for advertisers to [access that inventory] using programmatic targeting and frequency-capping techniques.”

In Google’s system, this inventory is exposed to buyers through a tool called marketplace, which also lists the prices publishers have set. Those avails are now coming to DoubleClick Bid Manager as well.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Actually – the fact that deterministic data is only 70-90% accurate is a myth that’s since been busted. Probabilistic approaches can be as accurate as deterministic if you look at the recent accuracy numbers that were confirmed by Nielsen – both Drawbridge and TapAd came in at over 90%, with Drawbridge topping out at 97.3%. Probabilistic is a legitimate, technological way to solve for cross-device identity. And it sounds like Google knows that, given their approach.

    • Ramesh — our scale rate was published with the Nielsen data announcement — basically tested against truth set that intersected with 10%+ of our graph — which is that precision at more than 120M uniques…our graph has more than 1B uniques — it’s hard to find a truth set/deterministic sample set of data large enough to see where that precision diminishes…I doubt FB will let us test our scale against their data.

  2. ramesh krishnan

    Still the accuracy rate is a myth , because the actual accuracy of a probabilistic model involves two metrics 1. Scale and 2. Precision. There hasn’t been any post regarding these two metrics in the accuracy rate published.

  3. Dave Eggars

    So if Google is using probabilistic cross device, why does Google prevent probabilistic cross-device from non-Google providers on its network?

  4. While it may seem that deterministic is leading in terms of accuracy, we have seen on many occasions that deterministic data isn’t 100% precise either.

    Imagine shared devices that will dilute the matching set (e.g. two deterministic profiles on one device), or low recall rates due to inferior cross-device deterministic sets. This can mean that probabilistic dataset can sometimes have higher precision rates than deterministic ones.

    In any case, it would be difficult to verify the Google dataset, as the chances of them offering it to third parties – beyond the Google stack – will be slim, if not non-existent.

  5. Kamal Krishnan

    Does DoubleClick’s cross device measurement is applied on the Data Transfer log files or is that a metric that is available only in the DCM interface?