Uncertainty is the new digital advertising reality

This announcement comes after a bunch of other new similar releases across the entire stack. You want to know how many store visits were driven by your ads and your website? Enter In-store visits, a Google Ads/Google Analytics report which estimates how many people showed up after clicking a paid search ad or browsing your website. Do you want to measure conversions across devices? Cross-environment estimated conversions is made for you.

Why is Google providing estimates, while digital advertising is supposed to be fully trackable and Google ubiquitous?

With the ever-more fragmented consumer journey across devices and channels and the new legal (GDPR/CCPA) and tech (ITP) constraints, marketers cannot rely on hard numbers anymore. No tracking system, even Google's, is able to measure the entire consumer journey with 100% accuracy. Ad exposures on Apple devices can no longer be stitched with the rest of the consumer journey, and neither can the vast majority of offline sales. Even the most basic consumer journey starting with a mobile search and ending with a desktop ecommerce transaction cannot be measured in most cases.

Because brands now need to measure performance across channels and because regulators and technology leaders are cracking down against cookies, fully deterministic measurement is dead and all your traditional reports are even more irrelevant every day. The only way to measure performance across devices and channels in a future-proof, privacy-compliant way is to use a sample and extrapolate. That’s why Google is doubling down on machine learning applied to cross-device, in-store and ITP-proof measurement.

The calculation is specific for each feature, but the methodology is always the same.

Google uses an undisclosed, supposedly huge sample of users for which they have collected the right consent first and they can accurately track across publishers, devices and offline channels. Then they extrapolate to the entire population, applying machine learning or more traditional online polling. The end report is always aggregated and only available above a certain volume threshold. Sounds like an old school Nielsen measurement? Well, it is panel-based digital marketing on steroids, leveraging Google’s enormous identity graph and ubiquitous app tracking capabilities. And because the identity graph is huge and data collection is in real-time, the final estimate is not only accurate, but most of all, granular and actionable - the two main limitations of traditional panel-based systems.

The question is not to take advantage of these new features. It is a no-brainer you should use them all day, every day. The question is much more how you integrate those new reports in your workflow and decision-making process and how to best change mindsets when measuring performance. Most of the features above are still in Beta, not always easy to export, and pretty much never compatible with third party systems. It makes it harder to incentivize a media agency on cost per acquisition, when the conversion is by definition estimated and not auditable. It also makes the Google Marketing Platform even more mandatory for brand marketers, and the question of Google playing the judge and the jury even more prominent.

However, there is a new reality to deal with and we at fifty-five believe no serious CMO should prioritize Google’s independence and existing agency routines over marketing performance in the long run.

Hugo Loriot, partner and managing director US at fifty-five.