Google ditches last-click attribution in favor of machine learning-based model
In an effort to provide more accurate, precise and privacy-centric measurement to marketers, Google has announced it is updating its ad attribution model. Instead of relying on consumers' last interaction alone, Google Ads will employ machine learning to assess data from throughout the conversion pipeline. The move represents a shift away from what’s commonly known as “last-click” models of attribution and furthers the tech giant’s investment in consumer data privacy.
Google is introducing a new attribution model
Google has announced it is updating its attribution model for marketers.
The company will no longer rely on last-click attribution, but will shift to what it’s calling “data-driven attribution," per a blog post published earlier today by Google Ads’ vice president and general manager of buying, analytics and measurement Vidhya Srinivasan.
While last-click attribution measures which touchpoint a consumer engaged with last before making a purchase, Google’s new framework employs machine learning to gauge everything from how conversions are measured to how to improve automated bidding in the media buying process.
While Google's ads business already offers this data-driven attribution model, it was not previously accessible to all advertisers, due to minimum data requirements as well as some limitations on types of conversion. Per the company’s announcement today, minimum data rules will be dropped and data-driven attribution will be made available to all advertisers in Google Ads beginning in October.
The decision to switch to data-driven attribution, Srinivasan explained in today's post, was informed by the changing privacy landscape. Consumers are increasingly demanding data privacy protections from big tech companies, leading the Googles and Apples of the world to introduce new policies and tools that give users greater say over how their personal information is used.
Unfortunately, many of these changes, including Google’s Federated Learning of Cohorts (FLoC) — its answer to the demise of the third-party cookie — and Apple’s AppTrackingTransparency (ATT), make it more difficult for marketers to understand user-level behavioral patterns, connect with audiences in personalized ways and, crucially, measure the impacts of their efforts. Ultimately, new frameworks like FLoC and ATT inhibit marketers’ ability to measure attribution and adapt their media, marketing and sales strategies accordingly.
Google argues that its new data-driven attribution is a means by which to preserve consumer privacy while giving marketers the tools they need to measure. Last-click attribution, on the other hand, Google says, is becoming increasingly ineffective on both fronts.
How data-driven attribution works
Data-driven attribution assesses signals throughout the entire customer journey — rather than the last touchpoint alone — and offers a more obfuscated view of user-level data, thereby offering improved privacy, according to Google. At the same time, the company says the new model has the potential to improve advertising effectiveness, since it analyzes all “relevant data” about the interactions leading to a conversion. The new machine learning-powered model evaluates everything from the period of time lapsed between interaction and conversion to ad formats. Plus, Srinivasan wrote, Google Ads evaluates results from holdback experiments to optimize the accuracy of its models.
The company claims that, “When combined with automated bidding strategies, data-driven attribution can drive additional conversions at the same cost-per-acquisition,” due to the model’s ability to “better predict the incremental impact a specific ad will have on driving a conversion, and adjust bids accordingly to maximize...ROI,” per today’s announcement.
In a quote shared in today’s post, Lara Harter, head of online marketing at DocMorris, said that Google’s data-driven attribution led to “an 18% reduction in cost of sales over last-click.”
The new attribution model, which is already available on Google Search, Shopping, Display and YouTube ads, will be expanded to support additional interactions, including offline as well as in-app conversions. The update will roll out as the default model across Google Ads beginning next month, but the company will still allow users to select from five different attribution models.
Plus, per a statement shared with The Drum, the company will soon integrate data-driven attribution into Google Analytics 4 and plans to introduce additional privacy-focused ad measurement tools and products in the coming weeks. Among these updates will be expanded access to Google Ads’ enhanced conversions and engaged-view conversions that enable marketers to measure conversions that happen within the days after viewing ten seconds or more of a given advertisement.
Google declined a request for comment on this story.