Google says that the new tracking tool is going to be available in Germany, Great Britain, Japan and the United States. Cross-device conversion tracking will be available across both the search and display network as well as through DoubleClick.
They also report that thanks to the improved tracking, advertisers are identifying around 16% more conversions in the United States. Thanks to being able to find what Google is calling leaked conversions, advertisers are better able to create and maintain their advertising on a variety of platforms thanks to having a better idea of what converts.
However, although I think they’re moving in the right direction, I’m not entirely convinced that all of the reported conversions aren’t of the unicorn variety, e.g. entirely imaginary.
Google says that cross-device tracking works like this:
Cross-device and cross-browser conversions are estimated using aggregate and anonymized data from people who have previously signed in to Google. Based on this aggregated data, we create an anonymous, aggregate estimate of the number of cross-device conversions attributed to AdWords campaigns.
Like all other conversions, cross-device conversions uses what’s called “last paid click” attribution. This means that the conversions are attributed to the ad that was last clicked before people make a purchase, submit a lead, download an app, place a call, or otherwise take a valuable action.
Additionally, in their Calculating Estimated Cross-Device Conversion guide, they say that:
If we observed 6 book purchases from previously signed-in users across mobile devices, we might use the other signals available to estimate that the actual number of cross-device conversions is 10.
In other words, while some of the conversions Google reports are based on actual reality, some of them are basically guestimated. Google says that they will only report conversions if they are at least 95% sure that they are within 10% of actual conversion amounts. However, if they are very confident of their algorithm’s ability to create this data, they’re going to release report it.
This doesn’t mean that they system is inaccurate, but I think there’s way more room for error than Google is admitting to.