Key Takeaways
Treat ad dashboards as claims, not proof that advertising caused a sale
A skeptical opening. At a 2019 advertising conference, researcher Nico Neumann challenged the belief that precise targeting and machine learning reliably improve campaigns. Audits he presented found that consumer data were wrong for 10 to 85 percent of people, depending on the dataset. An auto insurance experiment even found an ad reduced sales, despite industry models predicting a large positive effect. The room offered no questions, then moved on to a talk about Instagram ads.
Correlation is not cause. A person who sees an ad and buys something may have bought it anyway. Neumann argued that standard analytics often credit advertising for outcomes it did not produce, and that broad, untargeted campaigns can sometimes offer better returns. His evidence challenges the industry's confidence, not the possibility that particular ads work.
The conference scene is more than a colorful opening. It shows how professional consensus can persist even when an inconvenient finding threatens the business model underneath it. The methodological issue is familiar from medicine: an observed improvement after treatment does not establish that treatment caused it. Randomized comparisons, including groups that do not receive the ad, can help separate persuasion from preexisting intent. Still, the argument needs that same discipline applied to its own claims. A failed campaign or flawed dataset does not establish that all targeted advertising is ineffective.
A blink-speed auction turns each page load into a global attention trade
An auction in a blink. When a page loads, an opportunity to show an ad can trigger real-time bidding, or RTB: an automated auction in which advertisers compete for that placement. The bid, selection, and delivery take less than 100 milliseconds. These transactions happen at enormous scale, so people cannot evaluate each opportunity themselves.
Software runs the market. Demand-side platforms help advertisers set audiences, budgets, and bidding rules. Supply-side platforms help publishers offer space and set conditions for its sale. Their systems interact through ad exchanges, matching buyers and sellers. The machinery funds much of the free web, from search and social media to independent publishers. Its financial reach extends beyond media, since advertising profits also support other corporate investments and ventures.
The auction is not merely a faster version of a human sales negotiation. Automation changes what can be traded: tiny, momentary opportunities become liquid inventory available to buyers around the world. This resembles other digital markets where software lowers transaction costs and expands participation. But speed also makes meaningful inspection difficult. A system can execute a transaction flawlessly while the participants misunderstand what they bought. Efficiency in processing is not the same as reliability in value.
Standardize attention, and human experience becomes a tradable unit
From experience to units. Commodification is the process of standardizing varied things into comparable units that can be bought and sold at scale. In online advertising, an industry group called the IAB developed rules for counting ad impressions. One example, a viewable impression, requires at least half of an ad's pixels to remain in the visible browser area for at least one continuous second.
A lesson from grain. In nineteenth-century Chicago, standardized grades and receipts let grain from different farms be traded as equivalent goods. Buyers no longer needed to inspect each farmer's crop, and the receipts could be traded independently of the grain itself. Online ad standards work similarly: they make fleeting attention easier to price and exchange, even when the specific person, setting, or quality of attention is hard to know.
Standardization can make a market accessible and efficient, but it also defines what the market is able to see. A one-second, partly visible ad is measurable; whether a person understood it or found it useful is much harder to capture. This is a broader feature of administrative systems: what can be counted often becomes what organizations manage. The grain analogy clarifies how abstraction enables scale, though attention differs from a physical crop. It cannot be stored, inspected, or delivered in quite the same way.
More tracking data can hide a market that remains fundamentally opaque
Detailed does not mean transparent. Online advertisers can learn when an ad loaded, who was targeted, and whether someone later clicked. Yet this information may reveal little about the overall market: where an ad actually appeared, whether a fair price was paid, or why an algorithm chose that placement.
Three sources of opacity. Automated auctions make millions of decisions difficult to audit. Private marketplaces, or PMPs, let selected buyers access exclusive ad inventory, leaving others unable to see those transactions or prices. Major platforms also control much of the data used to judge their own performance. Facebook's push to sell more video ads was followed by the discovery that some video metrics had been overstated, including average viewing time by roughly 60 to 80 percent. Publishers had made costly content decisions based on those figures.
This is a useful distinction between data abundance and accountability. A platform may provide thousands of metrics while retaining control over the definitions, measurement, and access that give those metrics meaning. In measurement science, a proxy becomes dangerous when organizations optimize it as though it were the underlying outcome. The Facebook example shows that even a precise-looking metric can mislead decisions across an industry. At the same time, private deals can offer participants more direct information, so the challenge is not to ban every private arrangement but to prevent private information from distorting the public picture.
Ads lose value when real people stop noticing them
Attention is slipping. The first web banner ads, launched in 1994, reportedly drew a 44 percent click-through rate. A comparable display-ad average cited for 2018 was 0.46 percent. A click is not the only possible measure of advertising success, but the contrast captures how dramatically online users' responses changed. One Google report found that 56.1 percent of digital ads were never seen by a person.
Users push back. Ad blocking has spread across devices and regions. A 2017 study cited in the book found that three-quarters of North Americans regularly used at least one form of ad blocking. Younger users were especially likely to block ads or skip video commercials. The author acknowledges that brand advertising may shape attitudes without immediate clicks or purchases, but if an ad is unseen, it cannot have that effect.
The evidence is strongest as a warning about the gap between ad delivery and human attention, not as a complete measure of advertising's influence. Clicks capture some forms of response but miss delayed purchasing, memory, and brand associations. Conversely, advertisers can invoke those less visible effects so broadly that almost any campaign appears successful. The practical lesson is to match the measurement to the stated goal and test whether the result lasts. User avoidance also matters: people are not passive inventory, and repeated interruption can make the channel itself less welcome.
Digital attention can become subprime as prices rise while real value falls
A mismatch in value. Subprime attention describes the gap between the price of online advertising inventory and the quality of the human attention supposedly behind it. An ad placement is a claim on attention, not attention itself. It may be unseen, ignored, or delivered to someone who would have purchased anyway.
Fraud deepens the gap. Click farms and automated bots can generate activity without genuine interest. Domain spoofing makes an ad placement appear to come from a respected publisher when it does not. Estimates cited in the book put 2018 losses from ad fraud at $19 billion, while another analysis found that much display spending went to fraudulent or unseen inventory. These figures vary by method, but the central risk is clear: fake activity can make weak inventory look more valuable and encourage continued spending.
The mortgage comparison is useful because both cases involve assets whose apparent value depends on something underneath them. A mortgage-backed security depends on borrowers making payments; an ad placement depends on attention producing some benefit. The analogy should not be mistaken for identity: an ad is consumed immediately, while a mortgage is a long-term financial obligation. What transfers is the valuation problem. When buyers cannot inspect the underlying asset and sellers benefit from optimistic estimates, market prices can drift away from actual utility.
When ad budgets have fewer alternatives, weak markets keep attracting money
The money keeps moving. As newspapers and other traditional outlets lose advertising revenue, advertisers have fewer established places to put their budgets. Digital platforms gain both the audience and the money, while the decline of competing media makes digital advertising still harder to avoid. This resembles the savings-glut explanation of the mortgage crisis: a large flow of money searching for a home can inflate an asset that appears to offer a convenient alternative.
Middlemen profit from opacity. Agencies can buy discounted inventory and resell it to clients at a markup. Ad technology firms can add undisclosed fees or resell inventory passed through another intermediary. A study cited in the book found nontransparent agency practices were widespread. The participants who profit from this system have little financial reason to expose its weaknesses, even when advertisers and publishers bear the costs.
This account shifts attention from individual bad decisions to market structure. A buyer can keep spending not because every purchase is sound, but because the alternatives are shrinking and budgets must be allocated somewhere. That is a powerful feedback loop: money supports platforms, platforms attract audiences, and weakened competitors become less viable. Yet the analogy to the savings glut is suggestive rather than conclusive. Ad budgets are not identical to investment capital, and some advertisers may still shift channels when credible evidence of poor returns becomes available.
Advertising does not just fund the web; it shapes how people use it
Make people measurable. To sell attention efficiently, platforms need ways to identify users and record their actions. Profiles, likes, and other structured interactions make social activity easier to analyze and package for advertisers. The author connects this to legibility: making people and behavior readable to systems that administer or monetize them.
The interface follows the incentive. A platform organized around advertising has reason to favor actions that can be counted and targeted. That can narrow online expression toward recognizable profiles, reactions, and engagement metrics. The book imagines a more open-ended social space without accounts or fixed reaction buttons, but argues that such a space would be harder to finance through advertising. The point is not that advertising invented social interaction, but that its business requirements influence which forms of interaction platforms build and sustain.
The argument connects interface design to political economy: a business model helps determine which human activities become visible and valuable. This is a strong corrective to the idea that platforms are neutral containers. It also leaves room for other causes, including usability, safety, and the genuine convenience of profiles and reactions. The important question is not whether structure can be eliminated, but who benefits from each structure and which forms of interaction it excludes. Advertising can widen access to services while quietly shaping the terms of that access.
Require independent evidence and enforceable disclosure before trusting ad performance
Make claims testable. The book proposes a National Bureau of Advertising Research, an independent body that could test industry claims and publish evidence about ad quality, performance, and fraud. It points to two examples worth examining: Procter & Gamble cut about $200 million in digital advertising and reported a 10 percent increase in reach, while the New York Times stopped open-exchange buying on its European pages after privacy rules changed and reported no loss in digital advertising revenue.
Disclose what is sold. Drawing on securities regulation, the author proposes standardized information about ad inventory, including performance measures, business relationships, and conflicts of interest. Machine-readable disclosures could let software evaluate claims quickly. Legal responsibility for inaccurate reports would give sellers a reason to provide reliable information. Disclosure would not guarantee that an ad works; it would let buyers judge what they are purchasing.
The proposal combines two useful functions: independent testing challenges claims, while disclosure lets market participants assess what is being offered. This is more modest than asking regulators to certify that an ad is worthwhile. It also recognizes a familiar conflict of interest: platforms and intermediaries often control the evidence used to evaluate their own services. The examples are promising signals, not decisive experiments by themselves. Differences in audience, campaign goals, and timing can affect results, so an independent research body would need consistent methods and access to underlying data.
Plan an orderly transition before an ad-market crash harms smaller publishers
Collapse has collateral costs. Advertising helps pay for news, videos, newsletters, and services people use without a fee. If its revenue suddenly contracts, smaller publishers may fail before large platforms do. Consumers could face more paywalls, fewer free services, and reduced access for people unable to pay. The money also supports businesses and research beyond media, so the effects could travel farther than the advertising industry.
Do not confuse demolition with a plan. The author argues for a controlled reduction in programmatic advertising's influence, rather than passively waiting for the market to break. That requires alternatives and protections for those who depend on ad revenue. Otherwise, a crash could leave the same concentrated platforms in place, only with less independent media and more restricted access. The aim is not to preserve every ad-supported service, but to make change deliberate and less damaging.
The book's most provocative move is to treat market weakness as an opening for redesign rather than a reason to rescue the existing model. That makes sense only if transition costs are taken seriously. Advertising can fund widely accessible services, including work that cannot easily support subscriptions. Any replacement model should be judged by who can afford it, who controls distribution, and whether creators retain a path to audiences. A managed decline is therefore not simply a market event to accelerate; it is a public policy problem requiring choices about what deserves support.
Analysis
The book's central contribution is to shift the debate from whether online advertising is intrusive to whether its financial machinery is trustworthy. Its argument moves through a chain: software makes attention tradable at enormous scale; standards turn a fleeting human experience into comparable inventory; weak visibility into transactions allows prices and reported results to separate from actual value; fraud and avoidance worsen that separation; and intermediaries benefit from keeping money in motion. The 2008 financial crisis supplies a model for understanding these conditions, not proof that an advertising crash is inevitable.
That distinction matters. The book assembles evidence of low viewability, fraud, inaccurate targeting, and conflicts of interest, but it does not establish one comprehensive valuation of the entire ad market. Some campaigns do work, and clicks cannot capture every effect of brand advertising. Nor are digital ad budgets simply interchangeable with mortgage capital. The strongest claim is therefore about vulnerability: a market that cannot reliably verify its underlying value is exposed to a sudden loss of confidence, especially when its participants profit from optimistic measurement.
The analysis also makes the internet's funding model visible as a design choice. Advertising subsidizes access, but it rewards platforms for making users measurable and engagement countable. The result is not merely a way to pay for services; it is pressure on the shape of those services and the kinds of media that survive. This frames reform as a tradeoff among accessibility, privacy, creator income, and market accountability, rather than a simple choice between free services and paid ones.
The proposed response is strongest where it is restrained: independent experiments and enforceable disclosure can make claims contestable without promising that regulators can certify effective advertising. The proposed controlled demolition is less settled. Reducing dependence on advertising may open space for different institutions, but it could also accelerate consolidation or exclude audiences through paywalls. The book's durable lesson is to measure the system beneath the interface and to prepare alternatives before financial fragility turns into public damage.
Review Summary
Subprime Attention Crisis receives mixed reviews, with many praising its clear explanation of programmatic advertising's flaws and potential economic impact. Readers appreciate Hwang's comparison to the 2008 financial crisis and his accessible writing style. Some find the book's predictions compelling, while others feel the catastrophizing may be exaggerated. Critics note repetitiveness and a lack of nuance regarding major platforms. Overall, reviewers value the book's thought-provoking ideas about the future of internet advertising and its economic implications.
People Also Read
Glossary
Programmatic advertising
Automated ad buyingA system for buying and selling advertising inventory through software rather than individual negotiations. Advertisers and publishers set parameters in digital platforms, and automated marketplaces match bids with available placements. It enables advertising to be traded at very high speed and scale, while making it harder for people to inspect every transaction or placement.
Real-time bidding (RTB)
Instant ad auctionAn automated auction triggered when a website or app has an available ad placement. Buyers submit bids through software, the publisher's systems select a winning bid, and the ad is delivered as the page loads. The process typically takes less than 100 milliseconds, which makes direct human involvement impractical.
Commodification
Turning things into tradable unitsThe process of standardizing and abstracting something varied so it can be compared, priced, and exchanged at scale. In the book, advertising standards turn fleeting, uneven experiences of attention into countable units of inventory. The unit can then be traded without buyers knowing exactly who produced the attention or what its quality is.
Viewable impression
A partly visible adAn advertising industry standard for an ad that is at least 50 percent visible within a browser's viewable area for at least one continuous second after rendering. The definition helps buyers and sellers count whether an ad could be seen. It does not establish that a person actually noticed, understood, or was influenced by it.
Private marketplaces (PMPs)
Restricted ad exchangesPrivate arrangements in which selected advertisers can bid for a publisher's ad inventory, often under negotiated terms. They can offer participants more control and a less crowded marketplace than open exchanges. Because outsiders cannot see their transactions, however, PMPs can make overall supply, demand, and market prices harder to assess.
Subprime attention
Overpriced, low-quality attentionThe book's description of online ad inventory whose market price exceeds the value of the real attention behind it. An ad placement may be unseen, ignored, or exposed to bots rather than potential customers. Fraud and weak measurement can make this inventory appear more valuable than it is, widening the gap between price and underlying attention.
Demand-side platform (DSP)
Buyer-side ad softwareA software platform that helps advertisers purchase ad inventory through automated marketplaces. Buyers can specify audience characteristics, budget limits, timing, and other conditions, then allow algorithms to bid on suitable placements. DSPs can simplify large-scale buying, but their decisions and fees may be difficult for clients to inspect.
Supply-side platform (SSP)
Publisher-side ad softwareA software platform that helps publishers offer advertising inventory to buyers. Publishers can set auction rules, minimum prices, and restrictions on eligible advertisers. SSPs connect publisher inventory to exchanges and buyers, but additional intermediaries or undisclosed fees can obscure who is selling a placement and what the publisher receives.
Disclosure philosophy
Require sellers to reveal factsA regulatory approach, drawn from securities law, that requires sellers to provide accurate information about what they are offering without guaranteeing that the asset is worthwhile. Applied to advertising, it would require standardized disclosures about inventory, performance, relationships, and conflicts. Buyers would use that information to make their own judgments.
National Bureau of Advertising Research (NBAR)
Proposed independent ad research bodyA proposed organization modeled on the National Bureau of Economic Research. It would conduct independent research on the programmatic advertising market, test claims about targeting and performance, and investigate areas obscured by industry-controlled data. The book presents it as a possible counterweight to trade groups and platforms, not as an existing institution.
FAQ
What is "Subprime Attention Crisis" by Tim Hwang about?
- Core Thesis: The book argues that the digital advertising ecosystem, which underpins much of the modern internet, is fundamentally unstable and resembles the conditions that led to the 2008 subprime mortgage crisis.
- Market Structure: Hwang explores how online advertising has become highly automated, opaque, and financialized, drawing direct parallels to Wall Street and financial markets.
- Crisis Warning: He warns that the value of online attention is eroding due to fraud, ad-blocking, and declining effectiveness, creating a bubble that could burst with significant consequences for the internet.
- Call for Change: The book advocates for a "controlled demolition" of the current advertising-driven web, suggesting that reform or collapse could open the door to healthier alternatives.
Why should I read "Subprime Attention Crisis" by Tim Hwang?
- Understand the Internet’s Economy: The book provides a clear explanation of how digital advertising became the dominant business model for the web and why it matters to everyone who uses the internet.
- Reveals Hidden Risks: Hwang exposes the structural weaknesses and perverse incentives in online advertising that most users and even many advertisers are unaware of.
- Timely and Relevant: As debates about privacy, misinformation, and the power of tech giants intensify, the book offers a crucial perspective on the financial underpinnings of these issues.
- Actionable Insights: Readers gain not just a diagnosis of the problem but also ideas for how society might re-architect the web for the better.
What are the key takeaways from "Subprime Attention Crisis" by Tim Hwang?
- Advertising as a Bubble: The digital advertising market is overvalued and increasingly disconnected from the real attention and influence it claims to deliver.
- Opacity and Fraud: The system is plagued by a lack of transparency, rampant fraud, and metrics that are often misleading or outright manipulated.
- Systemic Risk: The collapse of online advertising would have ripple effects across the internet, media, and even philanthropy, given how many services depend on ad revenue.
- Need for Reform: Hwang suggests that only through greater transparency, independent research, and possibly regulation can the internet avoid a catastrophic crash.
How does Tim Hwang compare online advertising to the 2008 subprime mortgage crisis in "Subprime Attention Crisis"?
- Derivative Assets: Just as mortgage-backed securities were derivatives of real mortgages, online ad inventory is a derivative of real human attention, which is increasingly subprime.
- Market Opacity: Both markets became highly opaque, making it difficult for buyers to assess the true value of what they were purchasing.
- Perverse Incentives: In both cases, intermediaries (banks, agencies, ad tech platforms) had incentives to inflate the value of assets and obscure underlying risks.
- Potential for Collapse: Hwang argues that, like the subprime crisis, a sudden loss of confidence could trigger a rapid and damaging collapse in the digital advertising market.
What is "programmatic advertising" and why is it central to "Subprime Attention Crisis" by Tim Hwang?
- Definition: Programmatic advertising refers to the automated, algorithm-driven buying and selling of online ad space in real time, often through auctions.
- Scale and Speed: This system enables billions of ads to be placed every day, at speeds and scales impossible for humans to manage directly.
- Market Structure: Programmatic advertising has made the market more like Wall Street, with exchanges, dark pools, and high-frequency trading analogues.
- Source of Problems: The automation and complexity of programmatic advertising have introduced new forms of opacity, fraud, and misaligned incentives, making the market vulnerable.
How does "Subprime Attention Crisis" by Tim Hwang explain the commodification of attention?
- Standardization: Attention is broken down into measurable, standardized units (like "impressions" or "viewable impressions") to facilitate buying and selling at scale.
- Abstraction: This process abstracts away the context and quality of attention, treating all impressions as interchangeable commodities.
- Market Expansion: Commodification has enabled the explosive growth of digital advertising markets, but also made them susceptible to the same pathologies as financial markets.
- Negative Consequences: The focus on commodification has led to a system where the true value of attention is often ignored or misrepresented.
What are the main sources of opacity and fraud in the online advertising market according to "Subprime Attention Crisis" by Tim Hwang?
- Algorithmic Trading: Automated systems make it difficult to track where ads appear and why, even for industry insiders.
- Dark Pools and Private Marketplaces: Exclusive, non-transparent ad exchanges obscure pricing and inventory, similar to financial dark pools.
- Platform Dominance: Major players like Google and Facebook control key data and metrics, sometimes inflating or misreporting performance.
- Rampant Fraud: Practices like click fraud, domain spoofing, and fake traffic siphon billions from advertisers and further erode trust in the system.
How does "Subprime Attention Crisis" by Tim Hwang describe the declining value of online advertising (subprime attention)?
- Falling Effectiveness: Click-through rates and genuine engagement with ads have plummeted since the early days of online advertising.
- Ad Blocking: The widespread adoption of ad blockers, especially among younger and wealthier users, reduces the available audience for ads.
- Fraudulent Activity: A significant portion of ad impressions and clicks are generated by bots or click farms, not real humans.
- Demographic Shifts: The most valuable audiences are increasingly unreachable through traditional programmatic advertising channels.
What perverse incentives and conflicts of interest does Tim Hwang identify in "Subprime Attention Crisis"?
- Agency Arbitrage: Marketing agencies often profit by buying ad inventory at a discount and reselling it to clients at a markup, regardless of quality.
- Ad Tech Middlemen: Demand-side and supply-side platforms may collude to inflate prices, taking large, undisclosed cuts from ad budgets.
- Lack of Transparency: Both agencies and platforms have little incentive to reveal the true effectiveness or value of the ads they sell.
- Market Self-Interest: The profitability of the current system discourages meaningful reform, even as structural weaknesses grow.
What solutions or reforms does Tim Hwang propose in "Subprime Attention Crisis"?
- Controlled Demolition: Rather than trying to fix the broken system, Hwang advocates for a managed unwinding of the programmatic advertising economy.
- Independent Research: He calls for the creation of a National Bureau of Advertising Research (NBAR) to provide unbiased, rigorous analysis of the industry.
- Mandated Disclosure: Drawing from financial regulation, Hwang suggests legal requirements for transparency and disclosure in ad markets to restore trust and stability.
- Encouraging Alternatives: The collapse or reform of ad-driven models could open space for new, healthier business models for the web.
What are the broader societal implications of the "Subprime Attention Crisis" according to Tim Hwang?
- Threat to Free Services: If advertising revenue collapses, many free online services (search, social media, news) may become paywalled or disappear.
- Media and Democracy: The health of journalism and public discourse is tied to the advertising economy, making its instability a risk to democracy.
- Inequality and Access: A shift to paid models could exclude those unable to pay, increasing digital inequality.
- Shaping the Web: The dominance of advertising has shaped the very architecture and culture of the internet, limiting what kinds of platforms and interactions are possible.
What are the best quotes from "Subprime Attention Crisis" by Tim Hwang and what do they mean?
- "Attention is subprime. The bottom is falling out even as prices are pushed higher and higher."
This encapsulates the book’s thesis that the value of online attention is eroding, even as the market continues to inflate. - "The web that we experience...is sustainable only because of the continued health of programmatic advertising. Change that, and the whole edifice...begins to change with it."
Hwang highlights how deeply the internet’s structure depends on advertising, and how fragile that foundation is. - "The reality is that vast portions of the programmatic marketplace are not so much miracle cure as snake oil."
This quote challenges the hype around digital advertising, suggesting much of it is ineffective or fraudulent. - "There is, if anything, a strong ethical imperative to allow the collapse of global surveillance capitalism rather than attempting to save it, because it might clear the deck for something better to emerge."
Hwang argues that the end of the current system could be an opportunity for positive change, not just a crisis.
Download PDF
Download EPUB
.epub digital book format is ideal for reading ebooks on phones, tablets, and e-readers.