Home

Diminishing Returns

Online advertising’s dependence on surprise accelerates its own instability

When MTV News laid off a big chunk of its writing staff recently, the company line was that it would be “shifting resources into short-form video content.” When Vocativ fired its entire editorial team in June, it was “undertaking a strategic shift to focus exclusively on video content.” When Fox Sports 1 canned 20 writers and editors earlier this year, this was done, as Bloomberg reported, so it could move its “resources and business model away from written content and instead focus on our fans’ growing appetite for premium video across all platforms.” When Time Inc. axed 300 employees, it did so to “implement [its] strategy in key growth areas, such as video, native advertising and brand extensions.”

But contrary to the bromides from executives about what audiences really want, this “pivot to video” is more about advertisers’ ongoing search for a sustainable ad format, one that doesn’t end up proving its own ineffectiveness or exhausting its novelty. Americans appear accustomed to tolerating and even enjoying ads that resemble TV commercials — nearly four of 10 of them watch the Super Bowl more for the commercials than the game itself, according to a HuffPost/YouGov poll taken last year; and a 2015 National Retail Federation poll found that more than three-quarters of viewers consider the ads to be a form of entertainment themselves. And though Netflix and other services give viewers the option to watch commercial free, a 2009 New York Times article cited studies showing that people enjoy television shows more with ad breaks. This suggests that people view video ads as part of a broader social participation, and that they have become complacent about seeing them in the midst of other video material.

“Pivoting to video” is more about advertisers’ ongoing search for a sustainable ad format with inexhaustible novelty. Users eventually adapt to any approach for capturing their attention

It’s a different story with conventional online ads. People tend to experience them as irritating interruptions rather than a complementary form of content adding a sort of rhythm to media consumption flow. Unlike print or TV ads, online advertising often introduces an unwanted element of multimedia — overlaid boxes or autoplaying videos on an otherwise text-driven page. This aggressive approach — when some sales pitch blares at you suddenly while you were trying to read something — means online ads often have less in common with TV commercials and print ads than with sidewalk canvassers and telemarketers. Just as those nuisances might compel you to screen calls or change your walking route, disruptive ads are more likely to cause you to take defensive measures online. This undermines their measurable effectiveness and drives an attentional arms race.

According to web developer Maciej Cegłowski, “Advertising is like a disease: It takes people time to develop immunity and resistance.” Online users eventually adapt to any approach for capturing their attention or changing their behavior, which forces advertisers to continually modify tactics to remain viable. In 1994, a staggering 44 percent of people who saw the first banner ad clicked on it. Today, a “good” click-through rate is anything north of 0.05 percent. As users became accustomed to simple banner ads, developers made them blink and flash, added sound, and had them pop up over or under the window you were currently browsing. Then they began targeting users by demographic and behavior, an approach that has only become more complicated and invasive ever since. And from there, ads began infiltrating search results and social media feeds. Each new development came not necessarily because it would be more effective, but because the effectiveness of the earlier approaches had measurably waned.

Advertisers across media have tried to combat advertising fatigue by moving beyond traditional advertising spots, injecting themselves into conversations through aggressive social media engagement, viral stunts, and ads disguised as memes — think of Pepsi’s clumsy co-optation of Black Lives Matter and Burger King’s exploitation of Google’s voice-command technology. Similarly, more and more ads disguise themselves as “native content” in hopes of getting users to inadvertently pay attention them, or at least force them to spend mental energy differentiating between ads and non-ads. The distinction between the two gets blurrier on any ad-dependent platform. On a news site, for example, advertisements are beginning to look more like the articles, while the articles are increasingly discussed internally in marketing terms: clicks, time spent on the page, length of scrolling, etc. On the contemporary internet, it is difficult to find a form of content that does not serve mainly as an advertisement for itself, a whispered plea to keep engaging.

But there is only so much innovation possible within the realm of persuasive technique. Instead, the main engine of driving new advertising technology is the development of new metrics — new ways to monitor ad exposure and user responsiveness. “The more poorly current ads perform, the more room there is to tell convincing stories about future advertising technology, which of course will require new forms of surveillance,” Cegłowski argues.

So it is with the pivot to video. A pre-roll video before another video clip not only creates a more harmonious experience for the user, but it also fits more neatly into the framework advertisers are comfortable with from television commercials. People don’t click on video ads in great numbers (click-through rates for video are actually lower than they are for display ads), so ad brokers build their pitch around spurious “viewability” metrics and the value added through “brand awareness.” Even though users may not want more video content or more of their content published in video form, the advertising industry pushes publishers in that direction anyway. As a result of this unified effort, overall viewability of video goes up, and online ads seem again to be “working.”

But for how long? Will the internet just become an elaborate delivery system for television, or will the drive for measurable effectiveness break the system altogether?


The internet as we know it is largely shaped by the impact of advertising money. Before founding Google, though, Sergey Brin and Larry Page were wary of advertising. They co-authored a white paper that acknowledged advertisers’ interest in embedding themselves within search engines and offered a grave warning against the implications, arguing that “the issue of advertising causes enough mixed incentives that it is crucial to have a competitive search engine that is transparent and in the academic realm.” To retain its integrity, they said, a search engine would need to resist the temptation to let advertisers put a finger on the scale. Today, Google draws 89 percent of its revenue from advertising.

Positive interest is inferred from any sort of online engagement with products, and whatever desires and motivations can’t be tracked are presumed to be irrelevant. Dissatisfaction isn’t clickable

Google is not alone in this. Facebook, for example, draws 97 percent of its revenue from advertising. Countless other technology companies and media platforms rely primarily on advertising for revenue, too. Some host ads directly, others gather and sell data about users for advertising purposes, and the rest obtain venture capital in pursuit of future profit from advertising and data brokering. The companies that shape the internet depend on the $60-billion-a-year advertising industry to stay in business, which gives advertisers a leverage over our internet-dependent life far beyond banner ads. According to Fortune, Google and Facebook now make two-thirds of the online advertising industry’s revenue and account for about 90 percent of its growth. But what happens to their empires, and the internet we know, if digital advertising stops working? Or more precisely, what happens if they can’t come up with metrics that show that online advertising consistently works?

Though it’s clear advertising can work, precisely how it works is far cloudier. Department store magnate and marketing forefather John Wanamaker famously articulated this murkiness nearly a century ago: “Half the money I spend on advertising is wasted; the trouble is, I don’t know which half.” Measurement of traditional print and TV advertising’s impact is inherently imprecise — a matter of counting the potential number of viewers for an ad, and usually stopping there — and it’s inefficiently priced as a result. Comparatively, online advertising offers far more robust measurement opportunities. One can measure not only how many people see online ads but also how people respond to them and how many buy something as a result (an ad’s “conversion rate”). Trackers often know who specifically watched an online ad, and from where, and when, and how. The online environment in which digital ads appear is an arena of near total surveillance, where a wide range of behavior — from where one has been before to where one moves a cursor to where one rests one’s eyes — can be tracked and analyzed as ad-performance metrics and cross-referenced with other data about consumers’ behavior.

Wherever there are metrics, there are efforts to optimize for them. Advertisers continually tweak their approach as they chase higher numbers, A/B testing to see which shade of blue drives the most clicks and whether serif or sans serif fonts make people looser with their credit cards. But the result is not necessarily a greater ability to track what works, but an inversion thereof: consistently redefining “what works” in terms of what can be tracked, in a kind of self-reinforcing tautology. Metrics don’t uncover silver-bullet techniques of persuasion; they merely drive more and more elaborate metrics, more and more data collection. Positive interest is inferred from any sort of online engagement with products, and whatever desires and motivations can’t be tracked are presumed to be nonexistent and irrelevant. Dissatisfaction isn’t clickable.

This opens space for clickbait, “fake news,” native ads, and content created for “hate-reading.” None of this content serves readers’ interests, but it all reinforces the efficacy of metrics. Advertisers and their clients insist on this positivist approach, because the logic shores up metrics that otherwise look fairly dubious. Outright click fraud is also still common according to this report, and many “real” clicks may as well be fraudulent: about half of mobile ad clicks and as many as 75 percent of desktop clicks are accidental. According to a report from firewall maker Imperva, roughly half of all internet traffic is bots, which means a whole lot of what ad brokers present as clicks and readership is not human and has no potential to advance anyone’s business goals. There are potential conflicts of interest, too, with companies like Google and Facebook serving the ads and reporting their efficacy simultaneously. Many attention-capturing techniques involve some trackable user activity — clicking an X or answering a survey question or letting a pre-roll video play to completion — before users can resume what they were trying to do. This is less a proxy for engagement than for irritation. Basically, there’s a lot of noise in the data, and noise doesn’t buy tennis shoes.


The more intrusive forms of advertising metrics — personalized data collection about user behavior — are premised on hopes not only of targeting users with more tailored ads, but of uncovering which content activates the surrounding ads’ effectiveness for particular viewers. Whatever seems to correlate with making people look and click will be replicated, and that goes beyond the ads themselves to the content they support. The content is targeted too, not based on interest but on whether it is correlated to advertising engagement among our cohort. Content could effectively be optimized to induce a complacency toward ads. (This is one way of interpreting the content already on commercial television.)

As engineers tweak search and newsfeed algorithms, we are forced to consider any given piece of content not only in terms of what is about but also in terms of why we are even seeing it. It’s ostensible subject matter may be overwritten by what its algorithmic selection and the surrounding ads tell us about who we are. A seemingly mistargeted ad may make us question not only the algorithms but ourselves: Am I seeing an ad for a rehab clinic or a dating site for steampunk enthusiasts because of an accidental click or faulty code, or is it because I just haven’t been honest with myself?

Some degree of user grievance is tolerable if it ends up improving ad performance. Keep “pivoting to video” and watching the numbers, and disregard whether those viewers are humans rather than bots

Dynamically generated ads — orchestrated by ad brokers and programmatic software — also throw the identity of the sites on which they appear into doubt. Consider the once-popular “Around the Web” widgets that displayed spurious links to off-site content from otherwise reputable sites. Even if outlets publish creditable journalism, a user will likely associate their credibility with the deceptive links pushing reverse mortgages and nootropics below the fold. A ChangeAdvertising.org study from last year found more than 80 percent of the top news sites employed some sort of “content ads.” Only 46 percent of ads they hosted went to “legitimate” advertisers trying to sell products; 26 percent went to “clickbait” — posts designed primarily to display more ads, and rack up further clicks. This takes advertising logic to its absurd conclusion, in which ads serve to persuade audiences only to consume more ads, as if the only possible product were advertising itself.

The near future of online advertising seems to be expansion: According to statistics from the Interactive Advertising Bureau, the industry has grown by more than 10 percent every year but one since 2000, and Adweek cites projections that it will grow by about 16 percent this year. There are few viable alternatives to ad-supported models. News companies like the New York Times and the Wall Street Journal have seen some success from paywalls, but without a broad national focus or highly specialized proprietary content, the paywall model is rarely sustainable. Direct patronage models through platforms like Patreon and Kickstarter can work for smaller enterprises, but they typically rely on a level of personal affinity and trust that doesn’t easily scale. Even sites that aren’t directly dependent on advertising but still want visibility must optimize their content for the whims of search engine and social media algorithms, which are themselves optimized to deliver ads.

As online advertising expands, it will continue to seek novel forms — and novel engagement measures — that will reshape the environment in which we receive more and more of our information. Each new development begets a greater need for novelty to eventually supplant it. It’s unclear whether this can continue forever.

In the data-obsessed online advertising marketplace, video’s nebulous metrics give it power. Over time, though, video measurability will expand in depth and breadth and be combined with other data on individual audience members. Video ads will begin to be assessed for their measurable ability to affect individual behavior. There will be much more data, that will be much more inconclusive. And so there will need to be more new approaches and metrics.

Novelty is typically framed as offering some direct advantage for the consumer. A soda company, for example, might try to attract customers with a new flavor or a more ergonomic bottle. Novelty implemented on online platforms, however, often tends to alienate rather than attract users. How many times has Facebook redesigned its interface, highlighting some new feature and burying two others? How many Twitter renovations have sparked a predictable wave of outrage? Such changes often generate criticism and hostility from users, but that pushback is integrated into a larger calculation: Some degree of user grievance is tolerable if it ends up improving ad performance. Keep “pivoting to video” and watching the numbers, and disregard whether those viewers are humans rather than bots. Disregard whether certain tasks are better suited to text (like information retrieval and synthesis, informational depth, attention to nuance, and so on) than video. The metrics have spoken.

While the rate at which the internet allows us to consume content accelerates, so do the rates of exhaustion and disenchantment — the rate at which news gets old, trends become stale, memes lose their charm, or ads lose their effectiveness. Boredom seems to outpace the manufacture of content, but salvation feels perpetually within scrolling distance.

As an interactive medium, the internet can require a vigilance and alertness that other media do not. It expects us to be comfortable in shifting frameworks, adapting to various interfaces, navigating uncertain contexts, and adjusting our expectations and protocols of politeness on the fly. But given online advertising’s reliance on novelty in order to distract, capture, and manipulate our attention, it’s worth wondering whether our inundation with online ads is helping the industry reshape us according to how it wants us to be: outfitted with a helpless addiction to the new that perfectly suits a diminished attention span.

Adam Clair is a writer currently based in Philadelphia. He tweets infrequently at @awaytobuildit.