What happens if online advertising is just a big, fat bubble?
Oct 22, 2020

What happens if online advertising is just a big, fat bubble?

The profitability of targeted ads is a big reason why tech companies are constantly collecting so much data about us.

Big Tech monopolies are in the news this week. The Department of Justice sued Google over how it maintains its search dominance, and its search dominance is the key to its business model, which is that it makes 80% of its revenue from digital advertising. Facebook makes 99% of its revenue from advertising. 

The profitability of targeted ads is also a big reason why tech companies are constantly collecting so much data about us. And there’s a multibillion-dollar ad tech industry that exists because all of this makes so much money. But what if these ads didn’t actually work all that well?

I spoke with Tim Hwang, a former public policy executive at Google where he worked on artificial intelligence and machine learning. He’s the author of the new book, “Subprime Attention Crisis.” The following is an edited transcript of our conversation.

Headshot of Tim Hwang, a former global public policy lead for artificial intelligence and machine learning at Google. He’s author of the new book Subprime Attention Crisis.
Tim Hwang (Photo courtesy of Leafan Rosen)

Tim Hwang: I think this is an interesting and important distinction, which is it’s not necessarily to make the argument that advertising never works categorically — we have examples of it working. The question is whether or not the market as a whole really lives up to the promises that it’s made. And the promises made is that this data-driven, automated form of what’s known as programmatic advertising is a kind of advertising that’s way better than billboards or magazines, or the kind of “Mad Men”-style of advertising. And I think ultimately, it may just be that we are exactly where we were decades ago, which is that we actually don’t know which half of the money spent on advertising works and which one is wasted. It’s very difficult to tell.

Molly Wood: Is there an awareness of this? I know that after, for example, Cambridge Analytica, there was a lot of conversation about how there are lots of promises related to microtargeting and that it just might not be realistic. Are advertisers starting to understand this?

Hwang: Well, I think there’s a lot of willful blindness in the advertising space. This book that I just wrote, it opens on a really strange experience that I had going to a marketing conference where a professor laid out all of the evidence: 60% of people never see ads, ad blocking is up all around the world. And it was just dead space; the advertisers kind of refused to engage with it. It’s one of the things I’ve been thinking a lot about, because it’s similar to patterns that we see in other market bubbles, where there’s these deep structural problems with the industry, but a lot of the people involved either don’t want to hear it or they don’t believe it.

Wood: Listen, I’m very familiar with the idea, the belief, that technology must be working even with all evidence to the contrary. But I do want to ask you about targeting specifically, because it seems like there’s a lot of technical reasons it doesn’t work. But what about this idea that there’s a massive amount of data collection, that ads can be so specific and personalized that you literally cannot resist them? Are you saying that’s also not true?

Hwang: You brought up Cambridge Analytica earlier. There’s a fascinating report that just came out from the United Kingdom privacy regulator that was basically their research to say, “Look, there’s all of this kind of psychometric advertising. Does it make a difference?” And the conclusion there was no, there actually was not any significant difference. And there’s two reasons for that. One of them is that a lot of researchers find that a lot of the data being used is faulty and messy and doesn’t work. And I think the other one is whether or not this data actually helps you to target a message better is really unclear. There’s a great researcher by the name of Alessandro Acquisti who’s been doing some work on if you have targeted ads versus non-targeted ads, does it actually make a difference? And his conclusion is, it does, but really only by a small margin, much less than you’d think.

Wood: Could it also expose the fact that a lot of these companies no longer want the data for advertising? Like, they want it for machine learning?

Hwang: Yeah, I think that’s ultimately it. I think one of the great questions and responses I’ve had to the book is people say, “So why have we built this enormous surveillance infrastructure if this thing just doesn’t work?” I think people have traditionally thought, “Oh, well, it’s because Mark Zuckerberg wants to build a mind control ray; that’s his advertising system.” The reality is that data is being collected for other reasons. And for sure, I think things like the promise of machine learning is one of the reasons that people collect this data.

Wood: What can be done, do you think? I mean, this is a big, complicated technology question. You’ve got companies spending a ton of money and companies that rely on this for their whole business model. What could solutions even look like?

Hwang: One of my worries about this is that, again, if you study the history of market bubbles, a lot of what we see is very reminiscent to the, say, the subprime mortgage crisis of 2007-2008. And there is a momentum here, and the problem with bubbles is that while it may look great — in 2007, I think we’re saying how great the economy is doing — at some point they pop, and I think the human cost will be quite great. It’s really not just a matter of whether or not Mark Zuckerberg has fewer billion dollars. I think you got to think about all the media that’s reliant on this ecosystem, the journalism that relies on it, and the many other places that advertising touches online. I tend to believe in the idea that we have to find ways of deflating this bubble, so I’m really interested in the ability to both spread the public word about some of the problems in this marketplace, but also, I think there’s room for regulation. I think there’s room to enforce transparency in the marketplace to try to make sure that expectations about this match up with reality.

Wood: There’s been, just as a regulatory matter, there have been a lot of questions about banning targeted advertising. Should that happen?

Hwang: Yeah, I do think so. And I think in some ways it may be the thing that pops the bubble, because for the longest time, advertisers have been basically holding to the position that we need all this data in order to do our business, to target our ads. And what we’re seeing is things like the General Data Protection Regulation, the European privacy law, and the California Consumer Privacy Act, the California privacy law rollout, is in many cases, the market just keeps chugging along, even though advertisers have a lot less access to data. And I do think that that kind of realization, that all this data might actually not have been very meaningful, might actually cause expectations or perceptions about how great this stuff is to kind of crash to Earth. I do think these privacy laws have these two effects. One of them is to protect privacy, which of course is important, but I think the other side of it is actually it may strip the veil off this market that I think has been kind of shrouded for so long.

Related links: More insight from Molly Wood

How big a bubble are we talking here? Digital advertisers will spend over $140 billion in the U.S. in 2020. 

Speaking of artificial intelligence and Google trying to find possibly other revenue streams, The Intercept reported Wednesday that the company has a contract with Customs and Border Protection to help power the agency’s artificial intelligence work, including what’s being called a virtual wall between the U.S. and Mexico that uses drones, sensors and surveillance to find people trying to enter the country without authorization. This is, of course, definitely going to upset people working at Google who protested another government contract in 2018 called Project Maven that was shelved. 

As for that antitrust lawsuit against Google, AdWeek has a piece on how programmatic advertisers should probably be prepared for the possibility of that case expanding beyond its current focus on search ads. In fact, despite the fact that almost every state has some kind of investigation into Google’s business practices, just 11 states have joined the DOJ lawsuit — more will likely pile on over time. There’s been lots of talk about how Google actually controls pretty much every part of what’s known as the ad stack, and that’s rich soil to till for a busy state attorney general’s office.

Oh, also, Quibi shut down, about which I suppose enough has been said.

The future of this podcast starts with you.

Every day, the “Marketplace Tech” team demystifies the digital economy with stories that explore more than just Big Tech. We’re committed to covering topics that matter to you and the world around us, diving deep into how technology intersects with climate change, inequity, and disinformation.

As part of a nonprofit newsroom, we’re counting on listeners like you to keep this public service paywall-free and available to all.

Support “Marketplace Tech” in any amount today and become a partner in our mission.

The team

Molly Wood Host
Michael Lipkin Senior Producer
Stephanie Hughes Producer
Daniel Shin Producer
Jesús Alvarado Associate Producer