Skip to content

Predictions on Google’s predictive user-based targeting

The original beauty of Google Ads was that you could target intent-based searches…and you could do it without knowing anything about the person searching.

The searcher’s private information did not matter. Interests, friends, gender, income, none of it matters when someone is searching for Harry Potter boxed set or long HDMI cable or insurance for antique Porsche or Feldmesser 81 survival knife

At some point, this changed. 

I’d like to make a few predictions about the direction Google is going in, the reasons why they’re going in this direction, why it’s dangerous for privacy, and why it’s becoming a mess for Google. To note: I am not anti-PMax and I am not anti-automation. I am pro-privacy, and I have some very real qualms with the direction Google’s been heading.

But first let’s rewind and look at the fundamentals of the privacy issue.

What is privacy, after all?

So when I go to Google and type in “Feldmesser 81 Survival Knife” my privacy is not being violated. I am asking for information. I am granting this information about myself to Google. I am fine if Google assumes I may want to buy this knife. I am okay if I am served an ad where I can order this knife online.

In the past, ad platforms didn’t know whether I was a camping aficionado, an outdoors lover, a knife enthusiast, or a student of the blade. They just knew I wanted to buy the same survival knife that the Austrian Army issues to its infantry. Bingo. Here it is. Buy it.

Now, however, Google is building out a very detailed personas about you and I, and serving us ads on what it thinks we want (and it’s usually right, which is the potentially scary part). We know about a lot of ways that Google collects this data. Until Google claimed to stop this practice in June 2017, they scanned the content of Gmail emails to serve targeted advertising.

We know that Chrome links certain data to individuals, like location, search terms, browsing history, user identifiers, product interaction, and more. They even associate users with other users they interact with in real life.

Project Narnia is one of the weirdest things I’ve read about Google lately. Here’s an excerpt from the official Department of Justice complaint: “But in 2016, as part of Project Narnia, Google changed that policy, combining all user data into a single user identification that proved invaluable to Google’s efforts to build and maintain its monopoly across the ad tech industry. Over time, Google used this unique trove of data to supercharge the ability of Google’s buying tools to target advertising to particular users in ways no one else in the industry could absent the acquisition of monopoly—or at least dominant—positions in adjacent markets such as Search.”

Further, “The advertiser thus has no insight into how much the ad network spent to purchase a particular impression; the advertiser is charged a fee only when an internet user clicks on the ad. Google’s ad network, Google Ads, sets this fee based on the actual cost incurred to buy advertising inventory plus a markup. This prevents Google’s advertising customers from knowing how much Google is charging them, over and above Google’s costs, for the inventory.”

Google’s response to all of this has been an exercise in semantics for the past several years. 

From a Forbes article, “Google’s plan to replace cookies with so-called Federated Learning of Cohorts (FLoC), a clever way to say the “anonymization” of individual users into groups of individuals with common characteristics, is the kind of cleverness you’d expect from an ad giant.”

As the Electronic Frontier Foundation suggests, “Users and advocates must reject FLoC, and other misguided attempts to reinvent behavioral targeting. We implore Google to abandon FLoC and redirect its effort towards building a truly user-friendly Web.”

(Google claimed to abandon FloC, pretty sure they just renamed it Topics).

It’s clear: Google is moving away from targeting searches, towards targeting users.

The way they are harvesting data is problematic. But since they aren’t going to change their tech, they just change the way they talk about it. That is the entire reason they are moving to GA4: if they can’t explain a feature, they’ll just hide it. That’s the entire reason they’re pushing broader matches and hiding search terms in larger black boxes: if they can’t explain a conversion, they’ll just hide it.

They don’t want advertisers to know that a conversion came from scraping someone’s private communications. They don’t want users to think too hard about the fact that Chrome, Gmail, Search, YouTube, every single other Google property in existence…is collecting, fingerprinting, accumulating, analyzing, and predicting them. 

Sure, it’s fantastic when it comes to revenue…who is more likely to convert than someone whose friend just recommended a certain item?

Google’s now faced with a plethora of regulatory attention. They are increasingly operating on sort of a predictive “pre-crime” ad delivery strategy. It’s weird.

In an effort to appease the power that be, they’ve hurriedly renamed everything from user-centric to session-centric. Why else would “bounce rate” now be inversed “engaged sessions”?

What will happen over the next five years

I am going to make some wild, semi-informed predictions.

Google knows keyword-based search is going away, so they will make Google Ads keyword-less.

An increasing number of conversions in newer products like PMax are categorized into a black box of reporting (claimed to not be shown as a privacy measure, as in “search volume too low”) and this is actually because there are no keywords at all. Google is good at wordplay and semantics, so they will probably always call them keywords, but they will in reality be more like topic suggestions. These topic suggestions you add into campaigns will probably mostly be ignored.

Google will have a one-click ad manager for tiny advertisers

It’s been joked about for years, but it’s absolutely coming. There will be a big blue button followed by a credit card form. It’ll be targeted to the local clothing boutiques, the local handyman, the fledgling Etsy seller, that sort of thing. It won’t really affect agencies since we rarely work with this level of client anyway, but it’s definitely coming. I hope it performs well for these advertisers.

Google will target users not queries.

Everything privacy activists hated about Facebook, they should now hate about Google. As advertisers, we’re slowly being forced to be complicit in user-based targeting rather than query-based targeting. Google will still own lots of private user data, but advertisers will not have access to it. We will only be able to advertise based on what Google knows about their users. This algorithmic collection of data…this building of personas…is happening, has been happening, will happen, and is a privacy concern.

Google will severely restrict ad creative possibilities. 

As we’ve seen reflected in everything from RSAs to PMax to DSAs ad infinitum, lack of creative control will get worse. Advertisers won’t necessarily be able to write headlines or descriptions or custom banners or extensions. I think there will be some exemptions for compliance. But this is a big issue. We have larger clients who must have every creative approved by legal & marketing departments. And then, when everything is generated by ML, how does anyone truly stand out? Unless Google retains the ability for full creative freedom, advertisers will find other options.

Google’s reporting will be high level and generic, not detailed and attribution-focused.

We’re already seeing this in GA4, which reports at the highest generic levels possible. The problem is, this only works for very specific industries at very specific sizes. I don’t know how genericized, abstracted reporting will work for the vast majority of budget-sensitive small businesses, and even for larger B2B/industry advertisers. Not everybody is a large retailer spending millions on brand campaigns. That seems to be the only advertiser type Google cares about at the moment.

Will it work?

Yes, it probably will. Google’s machine learning is far from perfect. But it’s getting there. Once Google catches up with ChatGPT style natural language generation, you can bet your grandmother’s cigar that it will be implemented and it will actually start working fairly well. My issue isn’t with efficacy quite as much as ethics.

Do I have any proof?

Not great proof, but I have better proof than any rebuttal I’ve ever seen from Google.

The best proof I have is that Google’s personas are not perfect, especially if people are more privacy conscious (i.e. disabling cookies, or using encryption, or don’t sign into Google) and the ML can conflate users pretty easily. 

These “error” ads seem to be delivered most often on search partners, what I would consider low quality inventory, especially on fringe search engines like Brave. One interesting thing is that, from what I can tell, Brave does not notate advertisements with an “ad” symbol but cloaks it as an organic result (which is a Brave issue more than a Google issue). I believe PMax is heavily sending ads to search partner inventory.

For years folks wondered if messages were monitored and ads were served to them. It does sound pretty paranoiac, but time and again they’ve been proven correct.

I am aware of the Baader-Meinhof phenomenon. I see a lot of black Mercedes E550s on the road, not because there are a lot of them but because I notice every single one of them. 

Until we can do some sort of deanonymized test like Mike Ryan suggested, or a DOJ investigation forces something weird out, I don’t think we’ll be able to prove it. 

I am content with my gut belief on this one. I’m happy to be proven wrong. I would love to be proven wrong.

So what are our solutions?

The answer is not abandoning automation. For better or worse, we are forced to use it.

For some really good takes on PMax, I recommend Menachem Ani’s PMax thread(s), Ed Leake’s God Tier PMax video (you may not like the name, you may not like his face), and Kirk William’s articles on Google issues which I think are healthy critiques of this strange beast that makes advertisers a whole lot of money.

But I think there are some steps we should be considering.

Apple Search

Apple Search Ads for apps are absolutely fantastic. The ad platform’s user interface is fantastic. The instant Apple switches the Mac/iOS default search engine away from Google, is the instant massive market share goes away. Possibilities like this are probably why Google is trying to get away from search based ads anyway.

Microsoft Analytics

Clippy, are you out there? Will Microsoft Clarity evolve into something even better? Even if it’s not Microsoft Analytics, there will be some alternatives pop up which are far better alternatives than the absolutely crapfest which is GA4.

Legislation

I’m not too confident about legislation. I have doubts about whether GDPR or CCPA even does anything practically (Google still has all the data, they’re just calling it “engaged sessions” rather than “users”, right?) All the senators own Google shares anyway. But there is a chance pressure could do something. I will pay close attention to it one way or the other.

Starve the beast

Look, I owe Google a pretty successful career. But I don’t trust Google. I don’t use Chrome, I don’t use Android, I locked down our business Google Workspace accounts, I don’t use Google Maps, I use Signal for private messages, and I use Proton for my personal email. I am not sharing any of my real personal info with Google. If more people did this, it’ll starve the beast. The reality: 99% of people won’t do this.

Complain about it to Google

I am a de facto sales rep for Google. Our agency enables advertisers to spend millions of dollars annually with Google Ads. So I’m happy to pipe up and complain a bit. As an agency owner, we serve as the steward of client budgets, so we have an almost fiduciary duty to make sure they’re getting the absolute most out of their spend as possible. Individually, I think most Google reps are pretty decent, but they are also just as clueless as everyone else and usually have little direct real-world advertising experience. So it’s just Kool-Aid in, Kool-Aid out. The best thing we can do is hire them as soon as they quit Google, and then retrain. I’ve done it before, I’ll probably do it again. The good news is that Google’s various spokespersons are typically very receptive to feedback and pass on information. We’ve seen that happen in recent PMax updates.

Test other channels

Is another channel out there? Test it. Spend real money on it. If the beauty of Google’s automation is that it’s better at branding and awareness…do it manually and see how it works. 

In Summary

Let me answer some questions.

Am I overly paranoid? Always.

What is this stuff about the knife? The Feldmesser 81 is irrelevant, but it got me on a kick of testing personas and replicating searches. I’m tired of people telling me about the Baader Meinhof phenomenon. The point is that we, as users, are constantly being served ads based upon persona rather than search intent. And sometimes we only notice that when the persona is mixed up. Which happens often.

Why do I care? Because it’s my job.

Can I protect myself fully from Google’s privacy intrusions? No, not unless you retire to an internet-less cabin in Wyoming and destroy all of your technology.

Will my ads still work? Yes, ads will still work because Google wants us to spend more money on them. It’s just important to question how they work.

Will Google Ads specialists even be around in a few years? No. We will all be retired on the beach.

1 thought on “Predictions on Google’s predictive user-based targeting”

  1. Very interesting and different ideas you have put up in this article. Thank you for adding to our knowledge with these insightful points and everything is explained really well.

Leave a Reply

Your email address will not be published. Required fields are marked *