Google Ads announces machine learning-based data-driven attribution models in new privacy landscape

“[Data-driven attribution] uses advanced machine learning to more accurately understand how each marketing touchpoint contributed to a conversion, all while respecting user privacy,” according to an announcement from Vidhya Srinivasan, VP/GM Buying, Analytics and Measurement, Google Ads.

How it works. “Data-driven attribution looks at all the interactions—including clicks and video engagements—on your Search (including Shopping), YouTube, and Display ads in Google Ads. By comparing the paths of customers who convert to the paths of customers who don’t, the model identifies patterns among those ad interactions that lead to conversions,” says the about page.

Benefits according to Google. In its “About data-driven attribution” page, Google lists potential benefits for advertisers:

• Learn which keywords, ads, ad groups, and campaigns play the biggest role in helping you reach your business goals.
• Optimize your bidding based on your specific account’s performance data.
• Choose the right attribution model for your business, without guesswork.

The default last-click model only counts the final interaction toward the attribution, so advertisers have the potential to miss out on contributing micro-conversions along the user journey. “Data-driven attribution provides advertisers and businesses with reporting that better reflects the full marketing journey and higher performing bidding, which adapts to customers’ real journeys to conversion,” a Google spokesperson told Search Engine Land.

Ad availability. The data-driven attribution model is now available for Search, Shopping, Display and YouTube ads. The announcement also adds that Google will be “adding support for more conversion types, including in-app and offline conversions. We’re also removing the data requirements for campaigns so that you can use data-driven attribution for every conversion action.”

Opting out. For advertisers that do not wish to participate in the data-driven attribution option from Google Ads, the five rule-based attribution models will still be available:

1. Last click: Gives all credit for the conversion to the last-clicked ad and corresponding keyword.
2. First click: Gives all credit for the conversion to the first-clicked ad and corresponding keyword.
3. Linear: Distributes the credit for the conversion equally across all ad interactions on the path.
4. Time decay: Gives more credit to ad interactions that happened closer in time to the conversion. Credit is distributed using a 7-day half-life. In other words, an ad interaction 8 days before a conversion gets half as much credit as an ad interaction 1 day before a conversion.
5. Position-based: Gives 40% of credit to both the first and last ad interactions and corresponding keywords, with the remaining 20% spread out across the other ad interactions on the path.

However, it seems it will be a manual switch as “Over the coming months, we’ll be migrating existing conversion actions to DDA for many advertisers,” added Marvin. “When you set up a new conv action it’ll default to DDA (instead of last-click). You can select a different model if preferred. For existing conv actions, many will be migrated to DDA, but with plenty of warning to opt out & stick with another model. You can still switch any time,” she added in a new tweet. DDA will also be available in Google Analytics 4.

Other features and updates. Along with the transition to DDA, Google Ads has announced “a number of privacy-centric measurement features and product updates – many of which will directly impact advertisers,” said a spokesperson. These features and updates include the following:

• Enhanced conversions: As a follow up on our announcement earlier this year, enhanced conversions are now available to all eligible advertisers.
• Engaged-view conversions for display: A more robust non-click conversion measurement. Engaged-view conversions allow you to measure conversions that take place after someone views 10 seconds or more of your ad, but doesn’t click, and then converts within a set amount of days.

Why we care. Attribution has long been an issue that marketers have faced. This conundrum is especially salient as FLoC threatens to take away even more data from search advertisers — leaving them cobbling together data on their own. Google Ad’s machine learning attribution model seems to be Google’s solution to this lack of data. “Privacy-centric, DDA trains on real conversion paths & uses machine learning to measure and model conversion credits across touchpoints, even when cookies are missing,” added Marvin.

This is a “pretty big shift,” tweeted Kirk Williams, Founder of ZATO Marketing and PPC expert. “Data-Driven Attribution (DDA) was previously only available to accounts who had enough conversions in recent history to build the models to run DDA accurately.” To Williams, this indicates two big changes for Google:

1. All accounts can now run DDA immediately (I assume this means Google has enough confidence in its algorithms and sampling now, even for smaller accounts).
2. Attribution by default on accounts appears to have changed from last-click to DDA.

Many advertisers have claimed that the lack of data and reliance on machine learning makes their jobs harder (how can we optimize when we don’t know exactly what is causing success or failure?). This is another case where they will have to just trust the information that Google Ads is giving them without seeing the inside of the process. However, if done well, it could help many advertisers better understand which campaigns and ads are contributing to overall success throughout the funnel.