Anything that Lets Me See More of the World

There is a type of person who reinvests the dividends on their skill set, parlaying one nerdy experience into another, not so much raising the stakes as expanding the field of play, exploring whatever parts of the world make themselves available. When I think of this type of person I think of Tim Hwang, who among other things is the author of Subprime Attention Crisis: Advertising and the Time Bomb at the Heart of the Internet, which the SOW Members' Reading Group read recently. Among other things Tim is also a fellow at a think tank, a partner at a boutique law firm, and ostensibly my erstwhile/absentee boss at the Infrastructure Observatory.

In short Tim is extensively learned, and when we with him over the summer I honestly wasn't sure where the conversation would go. In the interview below (which has been edited for readability), we discuss the structure of the internet advertising industry, how to get a tour of a nuclear power plant, and USPS' rules for media mail — as well as Tim's philosophical approach to his absolutely fascinating career path.

Spencer Wright: I wanted to start by asking why you wrote this book? I originally got to know you as somebody who wrote a field guide to shipping containers, and who ran infrastructure tours in his spare time. How did you get excited enough about programmatic advertising to spend a year or two of your life writing about it? 

Tim Hwang: So I had the good (or bad) fortune of working at Google for a few years; I briefly ran public policy for them on AI and machine learning. This is in, like, the heady days of the late 2010s — before the last round of AI, pre-AlphaGo. It was a fun time, and I think one of the funniest things, which kind of inspired this book, is that I sat down with someone who worked at the company and I was like, "So I'm trying to find someone who can explain how we make money." And this engineer was like, "Oh, well, we make money through ads, of course." And I said, "But, no, really, how does it work? I just don't understand how the dollars go from someone else into Google's pocket." And this engineer, who works on self-driving cars, and all this next generation stuff, was like, "I genuinely don't know." It was so striking to me, because here was someone who was there at the company, arguably one of the biggest ad companies of all time, who basically had zero concept of how the technology actually worked.

For a lot of the folks I knew outside of the company, the fact that Google worked on advertising was known, but the details of it were very obscure. And so I got very, very interested, kind of for the same reasons I had gotten interested in doing infrastructure tours — it was just like, this is a fundamental piece of the way the modern internet works. It was odd to me that there wasn't a very clear explainer for how it all went down. This was the original idea of the book, but as I went digging, I started wondering, "Wait, is programmatic advertising actually a real business at all?"

SW: It's funny to think of this as an infrastructure problem. And maybe you should explain The Infrastructure Observatory, which you and I kind of worked on, in parallel, in the past.

Sure — so, the infrastructure tours: I lived in the Bay Area for about a decade. And of course the Bay Area has one of these major container terminals, just huge ships coming in and out all the time. And at the time, me and a friend were like, "Oh, it'd be super cool to get a tour of the container terminal." And so we messaged them, sent an email and they basically replied, "No, we don't do tours." And my friend and I put together an official-looking website, and we gave ourselves a special domain name, and then we emailed them back again. And this time we said something like, "Oh, I'm director of the Bay Area Infrastructure Observatory; can we have a tour?" And the guy got back to us and was like, "Oh, yeah, totally." 

So we ended up touring the shipping container terminal. We toured the Bay Bridge before it opened up. We visited a nuclear power plant. I don't even know if you're allowed to do these sorts of things now [Editor’s note: In my experience, you’re mostly not 🙁]. And the inspiration for it was, like, take the Bay Bridge: You drive across it, and you don't really think about it. But then to go out of your way on a weekend with a group of people to take a four-hour tour of the bridge… I think it gives you a much deeper appreciation of just how much work everything is. And in some little way, that feels important to me.

And so, yeah, I would say that the ad infrastructure stuff in Subprime Attention Crisis is very much in the same spirit — it’s like a bridge, or the power outlet in your home. You just browse around the internet and basically everything you do there is shaped by the incentives inherent in advertising. But we don't think about it on a daily basis; it's just the water in which we swim. And so the idea of writing this book was very similar to the Infrastructure Observatory: Let's just, for a moment, delve into what's going on here, and just appreciate how crazy of a system it is.

SW: This is a very effective analogy for me. The internet doesn't exactly run on advertising in the same way that my local grocery store runs on a shipping terminal, but advertising and shipping terminals are critical subsystems behind services that I really, really care about. I care about my access to groceries, so I should care about shipping terminals. Similarly, I care about my access to The New York Times — so I should care about programmatic advertising.

Yeah, I think that's right. I think one of the most interesting things about learning more about infrastructure is connecting the dynamics or incentives or developments that are happening in an obscure domain to, like, your everyday experience buying groceries. So, for instance, it turns out that the weird vagaries of international trade, and how expensive it is to move something across the Pacific, will have a huge impact on what you see on grocery store shelves. There's this great book, it's called The Fish That Ate the Whale, which is about the history of United Fruit. Basically, it's about how transporting bananas has been really, really hard traditionally. And our ability to solve that logistical problem has made bananas a lot more available to people.

SW: Throughout Subprime Attention Crisis, you make an analogy between the programmatic advertising industry circa 2020 and the mortgage industry circa 2008. How do those two systems compare? 

So you had this financial asset in 2008: the mortgage. People were very familiar with mortgages, but the mortgage itself got wrapped up in this bigger bit of financial infrastructure, and the result was that the heart of the US economy was just up in this completely strange financial engineering. There’s similarity there with the programmatic ad industry: We're familiar with a banner ad, we're familiar with a billboard. But the economy that that ad is now a part of is its own weird beast. Basically, things happen in one domain (programmatic advertising), and they have really strange downstream effects in our everyday experience of the web, which we might not otherwise accept or expect.

A friend was making fun of me about this book when it came out. He was basically, like, “normally you pick a metaphor that is simpler than the thing you're trying to describe.” But collateralized debt obligations are famously complex and obtuse. It's actually still unclear whether or not it was a good metaphor for the book. But it was helpful, at least to me. [Laughs]

Jay Williams: I take your point on commoditization: Just because you understand pigs, that doesn’t mean you understand pork futures; the commodity becomes a different thing. But even though online advertising has been commoditized, we don’t really see ad futures being sold, and I intuitively feel that there's a big difference between a market that contains futures contracts and one that doesn’t.

Yeah, I think that's right. I mean, look, I don't think there's actually anything wrong with commodification — turning an asset into something that can be traded in the marketplace. There are a lot of benefits to doing that. But I think the additional question is, like, what are incentives in a commodified market that would make it more or less sustainable? 

Commodification can create opportunities for markets to become pathological. So I think when you look back at the mortgage crisis, you don't necessarily say, "Oh, the problem is that we commodified mortgages." The problem is that commodification created room for all of these bad incentives; it resulted in a kind of pathology. So, maybe, commodification is a necessary but not sufficient condition for market failure.

SW: So, there's a lot of money tied up in programmatic advertising. And I think you built a pretty convincing case that a) programmatic advertising doesn't really work that well, and b) it creates all these opportunities for really bad incentive structures. And you argue, I think pretty convincingly, that the programmatic advertising industry is a bubble, and it's fairly likely to pop at some point. What are the stakes of that? 

I think they’re twofold. There's the most obvious one, which is that it turns out that lots and lots of media relies on advertising. So, if you believe in journalism, and the power of the fourth estate, then a lot of it just depends on programmatic advertising. And so part of the concern is, well, we've got this inflated bubble where everybody is selling garbage attention. And, like, do we have weirdly perverse incentives to keep this engine rolling? Because much of what we love about the web is powered by programmatic advertising. 

On a more fundamental level, though, I think a lot about how much of the web has been designed around advertising. I think the example that I put in the book is the “like” button. Why does the “like” button even exist on the internet? Well, one reason is that it's very difficult to tell whether or not someone has legitimately liked a piece of content if they just look at it. The best way of recording users’ reactions is if there's a button where they can actually tell you. And so the whole reason you have a “like” button is because you have an ad economy that needs to measure how much engagement an ad got. We've created all these affordances in social media for that. 

And so I think the stakes are really twofold. One of them is that the content that's flowing through the web is largely paid for and subsidized through advertising. But then the other one, which is much broader, is that our whole experience and design of the internet has been governed by the financial structures of advertising. And so whether you like it or not, a change in the advertising market might have very strong downstream effects in the kinds of platforms that are sustainable online.

I think one way of thinking about this question is, okay, imagine an internet without advertising. Would we still have the Facebook newsfeed? Would we still have the algorithmic feed on Threads or Twitter? And I think there's one class of people, which I myself would kind of lean towards, who would say something to the effect of, "No, because the problem is that you want to drive engagement on your platform, and that’s partially driven by the incentives of advertising." And so the argument is, well, content algorithms are very attractive from an advertising standpoint, and so we had to invent them. In some ways, it's like a natural evolution of the ad economy.

I think there are some people who would say that on the other hand, algorithms make platforms a lot easier to navigate. They solve the discovery problem. This is an interesting debate, which I think reasonable minds can differ on: Which affordances of the internet are the results of advertising? I think this is a big question looming over the book.

James Coleman: One of the things you discuss in the book is how as an advertiser, I buy attention — but a lot of that attention may not actually be someone taking in my advertising content. To what extent do you believe that advertisers understand that, and value it accordingly? So, I think a lot about Apple’s App Tracking Transparency policies, which they made a few years ago and which caused Facebook and other companies to lose a boatload of money. So Apple makes a change, the value of Facebook’s advertising inventory goes down, they lose billions — how does that make you think about the ideas about attention in the book? 

I think there's two things to be said about that. The first one is that we're in the midst of an interesting natural experiment. Prior to Apple making these moves, I think a lot of companies would've said, like, "We desperately need the ability to do all of this tracking in order to deliver effective advertising." But now that Apple has shut that down, what's interesting to me is that, sure Facebook lost a boatload of money, but you know who are still some of the richest companies in America? These advertising companies, these social media platforms. 

So there's one way of looking at this, which is, well, maybe we didn't really need all of that tracking and targeting stuff anyways. These platforms never had to collect this data. This is the idea behind contextual advertising. It turns out that if you're Googling around for plumbers, you're probably looking to hire a plumber. And that actually is all the data that you need in order to do the advertising that you need to. And so this dream of hyper-targeted advertising — we find you just as you're about to do this thing, and give you the ad just at the right time — that may have always been an illusion. 

One point of view on this, which I think is sort of interesting, is that the ad platform always wants to sell the dream that they know the most about the consumer. So, the data might not really change advertising outcomes. What the data does is persuade Coca-Cola to buy ads. So, maybe the fallout from Apple’s App Tracking Transparency reflects not necessarily any change in the effectiveness of advertising; it has just changed the willingness to pay for that advertising.

One funny anecdote: I have a friend who is in the ad industry, and when the book came out they told me, "Tim, you've got it all wrong. Advertisers don't buy ads to sell products. All they really buy ads for is so they can look good to their boss and then get awards at the Cannes ad festival." So, like, how much advertising is purchased just to convince your co-workers that you’re doing a good job? And that's a whole another aspect of advertising that has nothing to do with effectiveness.

SW: I wonder if you could talk a little bit about other ways that media is funded on the internet. At some point after you wrote this book, you worked at Substack, which obviously offers publishers an operating model that is at least partially independent from advertising [Note that in the early days, Substack’s terms of service actually banned advertising completely; that appears to have changed since]. And more broadly, I mean, I think the idea of publishing businesses with subscription models has become significantly more popular since the book was published — maybe not more profitable, but at least it’s easier for some people to wrap their heads around.

Yeah, for sure. So, this is at the point of the conversation where some people will be, like, “You'll get micropayments on the blockchain and that will solve everything publishing!” I feel like I'm old-fashioned in that sense: Either you're the reader, and you pay for it, or someone else pays for it in return for advertising (media can also be funded via patronage, I guess). And so, yeah — there's room for innovation here, but I do think subscription is the main alternative to advertising. And the question for the web today is, can we replace the advertising business model but keep all of the good things that advertising provides to the web?

Scope of Work is supported by our awesome Members, and through support from:

SW: What are the good things that advertising provides for the web? 

I think one argument for advertising is that it raises access. A world in which you have to subscribe to everything is pretty difficult. If you can't afford to pay for the media, you're basically locked out of key bits of information. That’s the tradeoff with subscription. You say, "Okay. Well, you got to pay." All right, well, then anyone who can't pay is by nature excluded. Which might not matter if you're running, like, a niche pizza review blog, but you may be concerned about it if you're running a primary news source.

That being said, I think it's really exciting to see companies get big on subscriptions. When I was at Substack, we would talk about what the psychological price of media was. Part of the question is whether, when a consumer goes to a piece of media, do they think, "Oh, surprising, it's free." There's a battle happening over our norms — over how much people are willing to pay for media — and the fact that subscriptions have become so normalized gives me a lot of hope. Advertising has fixed the psychological price of all media at zero, and it's very hard to build businesses that ask for more than that at the moment; the ad economy has sucked the air out of a lot of other businesses that might otherwise be created. But in a world where the median expectation is that media costs more than zero dollars, a lot of stuff becomes sustainable in a way that isn't currently.

SW: I wonder how this analogy would work with books. Books are a form of media that in most cases cost something — unless you have a library, in which case they're free. But the existence of libraries doesn't make me expect that I can get books for free. If I go to a used bookstore and I get a book for ten dollars, that's delightful; it's like a wonderful little treat. Is there a world where the internet might look like that? 

I think part of the answer is how much you think the cat is already out of the bag. The case study that I think is useful here is the music industry. In music, you had this brief period of Napster, and file sharing, and everybody within a particular age cohort could get every song absolutely for free with very little consequence. I think the result of this was to set the psychological price for music to zero. This was the case until the rise of Spotify: We finally created a music listening experience that was so seamless, and had so much consumer appeal, that once again we were able to convince consumers to pay a price for it. That took a lot of work, but it produced a norm shift. 

In the book case, what's interesting is that you had a monopoly, Amazon, come in really early and say, "I know these are text files, but you're gonna pay eighteen dollars for them, and you're gonna like it." And so basically this transition from physical media to digital media happens, but everybody went into it with a different psychological price in mind. You bought books for your Kindle for fifteen bucks and you'd be like, "Oh yeah, that makes sense."