Numlock Sunday: Tim Hwang on the Subprime Attention Crisis

By Walt Hickey

Welcome to the Numlock Sunday edition. Each week, I'll sit down with an author or a writer behind one of the stories covered in a previous weekday edition for a casual conversation about what they wrote.

This week, I spoke to Tim Hwang, author of the book Subprime Attention Crisis: Advertising and the Time Bomb at the Heart of the Internet, which is out this week. Hwang’s book was the subject of a piece by Gilad Edelman in Wired that appeared in Numlock last week, here's what I wrote about it:

The market for digital advertising was $325 billion in 2019 and is projected to rise to $525 billion by 2024. The proceeds of that industry — ads are 80 percent of Google revenue and 99 percent of Facebook’s — fuel lots of other Silicon Valley ventures and valuations, but a new book calls into question the very fundamental premise of targeted advertising, comparing the extent of automation and fraud in the industry to the mortgage-backed securities market of the 2000s. Ad tech middlemen take up to 50 percent of all online ad spending, with the premise that the money goes towards better targeting promotions for consumers, but research has demonstrated that targeting can be less effective than random guessing. About $21.8 billion worth of ads every year are blocked, according to Adobe, and a 2017 study found that as much as 56 percent of all ad dollars spent in 2016 were lost to fraud or unviewable inventory.

This topic fascinates me because as a person who works in the media, I’ve clearly got some skin in the game here, and as a person who started a newsletter that pretty explicitly rejects ad-based sources of funding — again, thanks for subscribing to make that possible — I’m really interested in this idea that there’s some serious issues in the ad-supported model

We spoke about how advertising fuels the tech business, how lots of the splashy cutting edge research coming out of Silicon Valley was funded by ad revenues, and how an issue in the ad market could reverberate through the beating heart of innovation in the U.S.

The book can be found wherever books are sold, from Amazon to your local book store, and Tim’s on Twitter @timhwang.

This interview has been condensed and edited.

You've written a book called the Subprime Attention Crisis. That framing is very deliberately alluding to another subprime crisis that the United States had in the housing crisis. Why do you think the internet ad boom is potentially in peril?

It's interesting, I've been in technology for the past decade and a half, and despite all of the excitement around the latest products and services, it's kind of interesting how little attention is given to the financial engine that drives many of the biggest companies on the web. On a day-to-day basis, we're not hearing about all of the developments in the world of ad technology. I got to digging, and that's really the origin of the book.

What I discovered is the surprising thing is that if you hear the marketing copy around these ad systems they'll say, "We're data-driven. We use the latest AI, and we can really do effective ad targeting.” But, once you start digging, you start to see these interesting examples, data, and evidence that suggests that this kind of enormous engine that we've developed to fund the internet really might be built on sand in a lot of ways. I can get into more detail around it, but that's kind of the core of the book and that's kind of the core argument.

There's a line from the book that's been going around, which is that, again, people forget that Facebook and Google have a reputation as technology companies, but if you look at their balance sheets, they're advertising companies that make 95 to 99 percent of their money from advertising. Yes, they're making tech, but that's a side hustle to them.

Exactly. Having worked at Google for a few years, it is so interesting that the ad technology is even obscure to people who work at the company. It's kind of in the closet. It's not really in the day-to-day discussion of a lot of what's happening at the company. I think that breakdown is so interesting, I think it's really important to get to the core business model here just because I think it explains so much about how the web is designed and where it really might be going.

There's been a lot of decisions made about how we have built the web that have really been at the service of ad tech. I'd love to hear your take on what this ecosystem is and what exists that doesn't need to necessarily, if not for ads.

This is the line that I've been using, but I think it's a good one, which is that you can almost use advertising and the economics of advertising as a kind of lens for seeing almost everything on the internet because it is such a pervasive force of gravity that shapes the design of these platforms. I think it really goes to the things that we don't even really think of on a day-to-day basis.

The example that I always use is the "Like" button. Why do we have a "Like" button anyways? One really compelling reason to have a "Like" button is that it gives you a very crisp, measurable thing about what types of content people are engaging with and liking. It's incredibly valuable from the point of view of profiling people to deliver ads to them. It's also incredibly valuable as a way of measuring the effectiveness of ads. I think it's so critical — just this thing — that you see in Spotify, that you see in Facebook, everywhere on the web, its a very basic primitive feature that is actually informed by the demand from advertising. I think from the smallest thing to the biggest phenomenon, advertising is always kind of this force that's lurking in the background.

So many of our systems have been built in service to ad tech that people don't necessarily want to even participate in.

I think that's kind of the most interesting thing, and I think this is kind of a B-plot in the book, but I think it's worth considering there has been a lot of fear around the kind of influence of, say, micro-targeted ads in the last few years. This book takes a look at it and says, "Look, the evidence for this stuff making a major impact is actually really low." It may be that we still want to kind of criticize the companies and resist what they're doing, but we might want to base it on more than the idea that they have some data-driven mind control ray in Silicon Valley.

I loved how you also go into how, number one, half of all the money spent on ads goes to middlemen who, as you just mentioned, don't necessarily improve the material impact of the ad, and number two, how ad placement and ad fraud is like half of all ads displayed, so a small fraction of the money based on some of these estimates is actually going to getting businesses out in front of people's eyeballs.

That's right. One of the most fascinating examples from the last few years is from Procter & Gamble, one of the biggest companies and the biggest buyers of advertising out there. They basically decided they were going to cut digital ad spending by $200 million, no small amount of money. What did they find? They found basically no substantive effect on their sales. In fact, their advertising reach was slightly up that year. I think that's just so fascinating, which is: what is that $200 million even going towards if not actually to shape consumer behavior in some way that's even measurable? I think part of it is middlemen, and this real question about, "does it actually do anything?"

What are the stakes here? What is at risk? What is the potential peril if people start to ask some serious questions about what's going on in their ad space?

So, some people have argued about this on the book, and I do think it's an interesting debate to have, which is: if advertising is in crisis, who cares? Is Zuckerberg going to just be like a billion dollars poorer? No one cares about that. But I think the argument I try to make in the book is that modern advertising and the wealth that modern advertising has produced is just tied up in all the parts of the economy now.

For example, I spent the last few years doing a lot of work in machine learning and AI and thinking about the public policy of those technologies. The leading labs doing the most cutting edge work are at companies like Google and Facebook. It's not like those labs are making money. Those are loss leaders for the company. In some ways, advertising is subsidizing the most cutting edge scientific research out there. You have to think that actually there's this unexpected thing where, if the advertising economy is at risk, there's all these other things that have now become tied to the advertising industry through these technology companies that are also at risk.

I think that's worth keeping in mind. The other thing I'd point out, which is quite obvious, is that, the entire media ecosystem now sits on these companies as well. There's a lot at stake in terms of viability of news and media, which is dependent on the continued functioning of these marketplaces.

As somebody who works in news and media, I would like you to expand on that, please.

I think there's a couple of ways of thinking about it, but I think one important dynamic that we see playing out is essentially that traditional media companies are in competition with your Facebooks and Googles of the world with the programmatic advertising market. By and large, over the last few years, what we've seen is that people have bought the argument that the digital advertising that these companies sell is way better than traditional advertising or the advertising of smaller outfits.

What you see is really a lot of money, move from these media companies onto the big platforms. I think what we've ended up with is an ecosystem that's extremely brittle as a result. The COVID-related economic downturn proves that even some of the biggest media companies don't have the wherewithal to sustain two months of a downturn.

I would say the other dynamic too is that these companies — the people who control these programmatic marketplaces — they also control a lot of the really key information about where consumer behavior is. That leads to a lot of problems. Famously, if you remember from a few years back, Facebook was like, "Everybody's got to pivot to video. Everything's going to be on video. Everybody's watching video." It later turned out that they overstated how much people were engaging with video by like 60 to 80 percent. Now, we can debate over whether or not that was intentional fraud or negligence, but the fact of the matter is that they control the information. Everybody had to move with them. There's a lot of journalists that were fired basically on the basis of those numbers. There's also this kind of problem where I think the media is now actually a lot more dependent on these companies as part of the economics of ads.

Ads as a business is just a colossal amount of money that you don't need to do a ton of work to maintain — obviously, I'm very much glazing over some important work there — but either way, if you look at what's happened with the broadcast networks and how they've had to turn away from expensive, scripted programming that was subsidized by ads towards more reality style programming, which is cheaper, you can see the shift in ad money can deliberately impact what the fun stuff that they're also working on is so it's less "scripted television shows" and more "machine learning experiments."

I hadn't really thought about that, but I think that is an interesting take. We can think about not just who gets the money from advertising, but then what do they spend it on?

What else do you dive into in the book?

I think there's maybe two things that are interesting to dive into. I think one of them is that it actually turns out that you don't necessarily even need to debate on whether or not ads work to believe that ads are garbage. Let me put it this way: Google did a study basically a few years back, which basically found out that 56 percent of ads are just never seen at all. They're delivered, but they're wrongly placed on the screen so that the person that it’s delivered to actually never even sees it.

Now, we can talk about a world in which you see an ad or you don't respond to it, but this is a world in which half of ads, they never even appear in front of someone. What's interesting is that you can get caught up in this debate on "do ads work?" or "do ads not work?" but I think it is only a smaller piece of this much bigger picture, which is we have a huge amount of fraud. We have a huge amount of incompetence in how ads are delivered, and all of that also kind of undermines the value of what's out there. I think that's the first interesting point.

I think the second interesting point is what do we do about it, and where do we move on from here? What should the internet look like? And I think there's two things. One of them is I think this is the value in making the comparison to the subprime mortgage crisis or even just financial crises in general, is that some people I've already seen pop up online and say, "Look, this market wouldn't exist if it didn't work, and so the existence of the market proves that it's really effective." I think that's just a very silly argument because the history of financial crises give us all these occasions where you had these massive markets that people thought were totally bedrock solid but then turned out to be basically based on nothing. The subprime crisis is a total example of that.

I do think that, unfortunately, there's a lot of incentives right now for people to just keep building up the bubble, to just keep the ball rolling for as long as possible. I really worry about that, because the bigger the bubble grows, the more painful it is when it deflates. I guess one of the things I'm really interested in is are there ways of combining regulatory proposals and just downright publicizing of the problems in the market as a way of slowing it down and slowly deflating the bubble because my worry is that in the absence of some other force, it's just going to keep growing to the point where I think the eventual collapse will be very painful.

Again, we are in an election right now, a lot of what people are saying is like, "Ah, Facebook is going to change the map," but then, again, you kind of alluded to this, materially, it's not the ads that are really doing the business here. A lot of them aren't even getting in front of the right folks.


Yeah, exactly. I mean, there was a really interesting thing. Again, there's just all this news that's coming out all the time about this stuff, but I was really intrigued by I don't know if you saw this. The UK privacy regulator put out its report on the Cambridge Analytica scandal, I think one of the most interesting parts of that story — and a number of people really pissed about it — is that they were like, "We looked at the evidence, and actually, it turns out that this psychographic targeting stuff probably was super overblown. It's actually unclear if it had that much of an impact." Now, I think that there's two ways of viewing that. One of them is, "Well, we shouldn't worry about it because it doesn't have that much of an effect," but I tend to be in the second camp, which is like, "Actually, I mean, there still are privacy concerns. The fact that people can attempt to do this has its own kind of bad effects on democracy." But again, I think our worries around what, say, Cambridge Analytica did, should rest on their ability to somehow predict and manipulate people and really more based on basic things like privacy. I do think it's interesting to kind of see this conversation evolve, particularly among people who were critics of the tech industry.

It's not just the only people who have stakes in this are Procter & Gamble and Facebook. The deli down the block might want to advertise, and if they're getting ripped off because half the ads aren't seen and another half the ads aren't placed, then that sucks. That's a market problem.

I think that's the interesting thing is one of the thing you often hear from the ads industry is, "We're very transparent." There's almost a magical thinking around data where they're like, "Oh, you have all this measurability. This is the most transparent market in the first place.” But I think it's really difficult, and I make this point in the book, to get an overall sense of what the market for ads is doing. We basically have a world where we have really, really deep knowledge about particular ads, but it's very difficult to get a sense of how the market as a whole is evolving. You do hear cases where people are like, "Honest to God, I will swear the ads work," and they probably do in that case, but I think the debate is this is anec-data, what we really need is more transparency about what the market as a whole looks like.

I've looked into it, and I would like to know how much it costs to place a Facebook ad on average, and that information does not exist.

Totally. To be clear, I mean, I think it's very easy for this argument to be like, "Okay, we should just ban all ads," and to go to the question of the future. I would say I tend to be more moderate. The question is not whether or not we want ads or not. The question is whether or not you want ads to be the economic monoculture that's responsible for everything that's happening on the internet.

My humble proposal is: maybe not? Maybe we ultimately need a more diverse business model because I do think in a lot of cases, advertising has kind of smothered other business models that could emerge. Having lived out in San Francisco for about a decade, it's very common to hear from people being like, "I pitched this business to this investor, and this investor is like, 'Why is this not advertising? It's a model that we know that works.'" There's been this interesting pressure away from experimenting with models that could potentially be some kind of replacement to the revenue that we get from advertising and might lead to a much healthier web.

The book is out this week!


Subprime Attention Crisis, why the economic engine of the internet is built on nothing.

People can find it anywhere books are sold. Any last plugs that you want to get in there?

If you're interested in this kind of stuff, I talk about it on Twitter a lot. I'm @timhwang, The book is available on your local bookseller and on Amazon, so check it out.


If you have anything you’d like to see in this Sunday special, shoot me an email. Comment below! Thanks for reading, and thanks so much for supporting Numlock.

Thank you so much for becoming a paid subscriber!

Send links to me on Twitter at @WaltHickey or email me with numbers, tips, or feedback at walt@numlock.news