Subprime Attention: Will Ad Bubbles Collapse the Internet?

In this episode

Advertising fuels much of the Internet today. But what if programmatic, targeted ad don’t actually work? Vulnerable Book Club reads Tim Hwang’s Subprime Attention Crisis.


Ads, lies, money, videotapes, and the potential collapse of the Internet as we know it. Oh well, scrap the videotapes. The other things are bad enough already.

Hello and welcome to Vulnerable By Design, your friendly immersion into the things that make us vulnerable. I am Chris Onrust. Today it’s time for our Vulnerable Reads book club. What are we reading? It is Tim Hwang’s Subprime Attention Crisis: Advertising and the Time Bomb at the Heart of the Internet, which came out in 2020. Hwang’s book is all about dodgy dealings in the online advertising sector. Plus its fallout for the rest of us.

Snake oil

At the heart of Subprime Attention Crisis are two claims. First, a (to me) shocking diagnosis. And a second, a prediction. Hwang’s diagnosis is: by and large, programmatic advertising doesn’t work. Or in any case, that it’s nowhere near as effective as it is often claimed to be. As Hwang puts it diplomatically: “… vast portions of the programmatic [advertising] marketplace are not so much miracle cure as snake oil.”

Second, there is Hwang’s prediction. Hwang observes that much of the internet today is built by companies that get most of their money from selling advertising space. Think Facebook (also known as ‘Meta’), Google, Alibaba, Bytedance (who own TikTok), Amazon. But also this holds for news media such as BuzzFeed, or The New York Times. Hwang predicts that if programmatic advertising doesn’t work, and if advertisers realize that it doesn’t work, then if they have got any business sense, they will stop spending so much money on this. But if they stopped spending that money, then somewhere down the line, this whole theater of whatever internetty things they built with that money, might very well collapse. Told you, we were in for a treat today.

The question is, of course, what’s the evidence? Is it really that bad? Should we retreat to a log cabin somewhere in the woods?

What is programmatic advertising?

First things first, what exactly is programmatic advertising? Advertising, if we go by a technical definition, is producing an openly sponsored non-personal message in a public medium to promote a product service or idea. Programmatic in this context just means: automation. So programmatic advertising is the automated buying and selling of digital ad space. Which generally comes with bulks and bulks of tracking and data. And with the option to target (or micro-target) ads, to specific origins, or micro target ads to specific audiences. For example, people who live in Vietnam, and love coffee.

Programmatic targeted advertising is big. In 2021, it’s estimated that globally, people have spent around €140 billion on programmatic ads. That’s nearly a third of all digital advertising spend. And for some cases, such as for display advertising—which are those text, image, or video ads that come creeping up into websites, social media feeds or mobile apps—this can be as high as 72% of all digital ads bought this way.

Automated superpowers

Now, why is programmatic advertising so popular? Automation is is always efficient, isn’t it? And the option to target just gives you superpowers. For example, say that you are a business selling coffee grinders. You could go old style and show your message to anyone who wants to see it. Both to coffee lovers, and to those who wouldn’t want to be seen near that type of aromatic hot beverage. But what if instead you could just, very precisely, advertise to only those people who do like coffee at the exact moment when they are looking for a new grinder? As an article in the magazine Forbes put it: “Programmatic technology allows for more precise targeting (…). This means more of your money can be immediately shifted towards ads that drive results and [fewer] wasted ad dollars go to people who are unlikely to buy.”

Sounds pretty good, doesn’t it? Except there’s only one slight problem. It doesn’t actually work. At least according to Subprime Attention Crisis it doesn’t.

Now I will admit I was a bit thrown off by this point. If you have listened to the news at any point over the past decade, concerns about microtargeting were all over the place. Microtargeting supposedly influenced votes and election outcomes, most notably the presidential elections in Brazil in 2018, or India’s 2019 general election. For this reason, microtargeting has been called detrimental to society, or even a threat to democracy. Now, personally, I felt really clever, knowing about all of this online manipulation. But now it turns out that nothing actually happened? What’s going on here?

Well, luckily for us, Hwang doesn’t skimp on the evidence for thinking that something’s off with the programmatic ad sector. So let’s hear it.

Ad block & co

Why does Hwang think that programmatic ads don’t work? Well, first of all, many ads are actually never seen or heard. There has been a dramatic increase in ad blocking. An ad blocker is a piece of software that prevents an advertisement from appearing on a page. Ad blockers have never been so popular. So today, in 2022, to an estimated 42.7% of Internet users worldwide use ad blocking tools. In some countries, such as Indonesia and India, it’s even more than half of all internet users who block ads. One survey showed that people on the internet find ads: annoying, intrusive, they eat up screen space, they slow down page loading times and they slurp up data. Or, as a respondent said: “There are too many ads on the internet.”

Just to say the obvious. If someone blocks your ad, the ad will not load. People will not see or hear that ad. And if people can’t see or hear that ad, the ad cannot somehow magically persuade them to buy your product. That’s just … No.

Similarly, nearly a third of display ads turn out to load in places where people cannot even view them. They dangle somewhere down the bottom of a webpage where you didn’t even scroll towards. Again, if an ad isn’t seen or heard, it cannot persuade anyone to buy anything or to vote in a certain way.

What about the ads that are shown and that are viewable? Things don’t look too good there either. Click-through rates (which is the proportion of people who actually click on an ad data shown to them) have dropped dramatically. Last year, the average click-through rate for display ads was only 0.46%. So less than half a percent. That doesn’t sound like a whole lot to me. And let’s also not forget that quite a number of those clicks will have been so called ‘fat finger errors’. Where you didn’t even mean to click on anything, but the sheer size of your fleshy digits relative to those tiny buttons on screen activated something and you try to get back, but it’s loading something else and … Or is it just me?

Okay. Say that someone does see your ad, and then at some later point in time buys a coffee grinder. What does that actually show? Just because one thing happened after another, that’s no reason to think that it was because of the ad that the person bought the coffee grinder. If the sun rises after I’ve had my first cup of coffee, that doesn’t mean that it’s because of my cuppa, that the sun is going up. As far as I know, the sun cares terrifyingly little about what I’m up to these days.

Another fun fact, it turns out that ads are supposedly most effective on people who were already loyal customers anyway. That makes you wonder: If someone was already going to buy your coffee grinder anyway, why spend your precious marketing dīnār on advertising to them?

Joys of click fraud

Not to hammer home the point too much, but how about the joys of click fraud? Some of the clicks that make online ads seem so effective actually were not clicked by its intended audience at all. Instead, those clicks come from bots, or from people working in so-called ‘click farms’, workplaces where people are paid to repeatedly click on ads to make those ads look popular. Yes, we’ve ended up in a world where click farm employment is a thing. Let’s leave the ethics of that for another day, shall we?

Facebook got into a lot of trouble a couple of years ago because they had grossly exaggerated basically lied about how much time people were spending watching Facebook videos.

Hwang lists way more evidence than this, but I think you get the point. All this, Hwang says, makes programmatic targeted ads seem way more effective than they actually are. And that makes people want to invest, because it seems so effective. Which artificially drives up prices. Hwang concludes: “… ads capture an increasingly small portion of available attention online, and what these ads do capture is shrinking in real economic value.”

Time bomb

That is why Hwang calls programmatic ad inventory ‘subprime’. ‘Subprime’ being a term from finance, usually used to describe risky dubious investments or loans. And the parallel with the financial sector is not an accident here. Hwang wants to remind their readers how the bubble of risky subprime mortgages contributed to the financial crash of 2007 and ’08. Similarly, with programmatic, targeted ads that seek to buy and sell people’s attention, everybody’s cheering that they’re so super effective. But in reality, it is actually all hot air. And the entire castle may very well spectacularly collapse.

As I said, to me all of this ‘it doesn’t work’ diagnosis came as a bit of a surprise. But it turns out, Hwang is not alone in saying this. Other researchers are converging on the same conclusion.

‘It doesn’t work’

For example, in a 2019 public lecture, lawyer and philosopher professor Mireille Hildebrandt already pointed out that there’s just no proof that microtargeting works. And journalist Dr. Felix Simon, who has looked into this, says that ad tech companies present microtargeting as something magical. They make huge promises. But it is mainly hype, and people are buying into it.

And it’s not just the research is talking. A growing number of organizations have slashed their spending on programmatic, targeted advertising. So some years ago, Procter & Gamble withdrew around $200 million USD spending on this type of advertising. Similarly, Uber, cut back around 120 million they spend on this. And eBay did the same already in 2012. These companies, according to the reports, saw no change, no negative effects on their business. That suggests strongly that those ads weren’t doing much for them in the first place.

For me, all of this solidifies Hwang’s main point: That there’s good reason to think that on the whole, programmatic targeted advertising is just not working. So that is good news, in a way. Whatever other scary things are going on, here is one fewer thing to worry about.

But then there’s the bad news. On the whole, programmatic, targeted advertising doesn’t work. Hands up, if in the past week, you have watched a video on YouTube? Or did a Google or a Yandex search? You used Baidu maps to get around town? Or you relied on Facebook to remember a friend’s birthday? Maybe you still got a lone Gmail address lingering around somewhere? Oh, you’re actively using it? I see.

What if the bubble bursts?

Google, Facebook, and many, many more online platforms and news media get much of their money from advertising. For example, in the third quarter of 2021, Google got over 80% of their $65 billion USD in revenue from ads. For Facebook, in the same quarter, the percentage was over 97%. Facebook’s business model doesn’t rely on advertising. It is advertising. Facebook (or ‘Meta’) is an advertising business. It just happens to use a social surveillance platform to facilitate those ads.

What does this all mean? We have already seen that some ad companies are massively cutting back what they spend on programmatic targeted ads. What if other organizations will follow suit? What if organizations will en masse stop spending those billions on this subprime good? This is what Hwang calls a ‘bubble’ and a ‘time bomb’. If advertisers collectively started pulling out their cash, if a chunk of this ad tech revenue will fall away, are you sure you will still have access to your Gmail account? Are you sure that there’ll be some online map to get you to your destination? Will online news media still be around? Will you remember your friend’s birthday?

Hwang thinks that if the programmatic ad sector collapses, the result could be pretty disastrous: “… the internet we have, for better or worse, is yoked to the structure and prospects of the advertising economy. If this system of advertising is brittle, then the Internet as we know it is brittle.”

If you are now thinking: “Obviously that’s not going to happen. People will always continue to buy programmatic targeted ads. All will be fine.” Then, let me remind you that that’s exactly what the Dutch must have thought about tulip bulbs in the 1630s. And that didn’t work out so nicely then did it?

The whole point of there being a bubble is that something, however worthless it is, in reality, seems appealing enough for lots of people to buy into it. A bubble doesn’t seem like a bubble, until it actually bursts. And once it’s actually burst, it is already too late then, eh? I’m not a forecaster. So I can’t tell you whether a programmatic targeted ad bubble is going to burst in 2022, ’25, ’28, or even whether it will burst at all. What I can tell you, or rather ask you, is: Do you really want to wait? Hope that it will turn out all right?

Internet shapeshifting

Hwang’s book, Subprime Attention Crisis, is an alarm call. (Especially with that word time bomb in the title, of course.) But I actually like to see it as an invitation. An invitation to think about an alternative. What better moment is there for us than now, today, to ask: What do we actually want?

The Internet is only a couple of decades old. It wasn’t a given that it would take this particular shape. And—shock, horror—it certainly is not a given that it must continue in this in this exact shape forever. Clearly, if people are tired of ads, if people are unhappy with an ad-driven Internet where every single web page is spying on them, maybe let’s take it upon ourselves and see if, step by step, we can build something of an alternative that doesn’t, well, suck? Another world might, quite possibly, be possible.

Thank you for joining us at Vulnerable By Design this week. If you have a suggestion for the next for vulnerable book that we should read, please do get in touch. You will find all of our episodes and contact information on I’m Chris Onrust. Thank you for listening and bye for now.

See also