Real-Time Bidding: The Ad Industry Has Crossed A Very Dangerous Line
This post is a collaboration with Dr. Augustine Fou, a seasoned digital marketer, who helps marketers audit their campaigns for ad fraud and provides alternative performance optimization solutions; and Jodi Masters-Gonzales, Research Director at Beacon Trust Network and a doctoral student in Pepperdine University’s Global Leadership and Change program, where her research intersects at data privacy & ethics, public policy, and the digital economy.
The ad industry has gone through a massive transformation since the advent of digital. This is a multi-billion dollar industry that started out as a way for businesses to bring more market visibility to products and services more effectively, while evolving features that would allow advertisers to garner valuable insights about their customers and prospects. Fast-forward 20 years later and the promise of better ad performance and delivery of the right customers, has also created and enabled a rampant environment of massive data sharing, more invasive personal targeting and higher incidences of consumer manipulation than ever before. It has evolved over time, underneath the noses of business and industry, with benefits realized by a relative few. How did we get here? More importantly, can we curb the path of a burgeoning industry to truly protect people’s data rights?
The Evolution of Digital Ads
There was a time when advertising inventory was finite. Long before digital, buying impressions was primarily done through offline publications, television and radio. Premium slots commanded higher CPM (cost per thousand) rates to obtain the most coveted consumer attention. The big advertisers with the deepest pockets largely benefitted from this space by commanding the largest reach.
When digital came along, the advertising market exploded to now include ad impressions across online publishers. This had the potential to garner stronger visibility where consumers had quickly migrated their attention. Early publishers relied on traffic to command premium CPM rates in highly trafficked parts of their sites. For the publishers, their goal was to bring the coveted consumer eyeballs to their site. The more that came, the higher the impression levels, hence the greater the advertising revenue potential. Ad inventory was finite and for many of the early successful publishers like MSN and Yahoo!, they were not only able to garner high, consistent consumer traffic to their sites, they also dominated the online advertising for a time before Facebook or Twitter launched their own publishing platforms.
Advertisers traditionally would leverage their media buyers to take advantage of the best inventory, carefully negotiated across the entire ad buy to get the best CPMs across the publishers sites. Eventually, publishers realized that it would be more efficient and more profitable to build a network of publishers to understand and capitalize on consumers’ paths across the web. Over time, this finite inventory was replaced by the creation of ad networks with seemingly endless advertising opportunities. The long tail “promulgated the idea that collectively a large number of small sites could rival the scale of a small number of large sites”. The opportunity for smaller sites and blogs to monetize their own content suddenly democratized a space that was only accessible to the larger known publishers. It also opened a Pandora’s box that welcomed questionable sites into the network, questionable ad inventory levels and questionable ad performance. “The scalability of digital ad tech—that is, without the limits of the physical world—combined with the greed of its creators led to an explosion of supply (digital ad inventory) that far outstripped even the large increase in demand, as more dollars from traditional channels like TV shifted into digital.”
The Advent of Real Time Bidding (RTB)
Real Time Bidding (RTB) came into the market 2014 and effectively disrupted the media buying industry by allowing a much more efficient way to purchase advertising, bypassing media buyers, publishers and ad networks and using demand-side platforms (DSPs) to buy impressions among a marketplace of publisher sites, while targeting specific users. As an impression is loaded on a user’s web browser, information about that user and the page they are on is passed to an ad exchange, that allows advertisers to bid in real-time, for those impressions. By allowing advertisers to see the user profile and sites they’ve landing on in real time, the goal was to minimize the occurrence of wasted impressions on wrong users.
What has made this more enticing for the advertiser is the auctioning process that has exposed the advertiser to bidstream information about the users regardless of whether they were the winning bidder. This bidstream data reveals identifying information on individuals such as IP address, zip/postal code, GPS location, browsing history, device identifiers including device type, brand, model, screen size, connection type, OS and CPU speed. “Any bidstream data exchanged will be stored in an advertiser’s and publisher’s database until it’s manually deleted.”
The evolution of digital advertising has gone beyond accepted principles to find the right customer for the right product to outright surveillance of individuals.
The quest towards increased understanding more about users has surfaced myths surrounding the effectiveness of “long tail” niche sites, behavioral targeting and hyper targeting.
This really took off in 2012-13 when programmatic ad exchanges became more common. The exchanges needed to show buyers there was something new and unique—so they sold the idea of being able to target users at a much granular level. Behavioral targeting is based on the behaviors of users, for example, which sites and pages users visit. The ad tech companies would use this website browsing history to infer user preferences and characteristics. For example, if the non-logged in user visited ESPN, Sports Illustrated, Playboy, they were more likely to be male. However, over time, as the “nature and content of sites became more diverse, those assumptions and algorithms that were used to infer user characteristics became less accurate”. What’s more, the collection and use of this data only serves to artificially inflate demand—and revenue—for the digital middlemen to the tune of upwards of 500% more profit in exchange for a mere $0.000008 ROI per ad.
Hypertargeting is a related term which just means marketers are buying dozens of targeting parameters to use in their campaigns, instead of just using basic ones like age range, gender, demographic data, geographic data, and the like. Marketers pay extra for the delivery of super targeted ads but similarly, don’t get any lift—or value—in outcomes. So, it too is a waste of money. But because marketers have drunk the kool aid, more ad tech companies compete to collect more data so they could sell more targeting parameters—like buying data from credit card companies to see what things people buy to infer the other things they are likely to buy in the future.
“Marketers continue to believe these myths and increase the ad budgets they spend on it; but as far back as 2010, groups like the EFF (Electronic Frontier Foundation) were already sounding the alarm about “tracking without consent". They released a demo called Panopticlick in 2010 to show consumers what and how much was being tracked, even if they deleted cookies—that is, the consumer has no choice or recourse.”
Despite these Myths, Demand does not Show Signs of Waning
Almost seven years since its introduction, the global Real-Time-Bidding (RTB) market is expected to grow from USD$6.6 billion in 2019 to USD$27.2 billion by 2024, at a Compound Annual Growth Rate (CAGR) of 32.9%. Post Cambridge Analytica, which exposed the harvesting of Facebook user profiles to manipulate voters at the ballot, the notion of Surveillance Capitalism, coined by author Shoshana Zuboff perfectly sums up this market characteristics as the “unprecedented asymmetries in knowledge and power that accrues to knowledge.” Similar to the growth of RTB, The Online Advertising Market was valued at USD 304.0 billion in 2019 and is expected to reach USD 982.82 billion by 2025.
The alarms that the Electronic Frontier Foundation(EFF) and others have raised about RTB would be brought to bear. What has transpired, as a result, is the culmination of user information collected by hundreds of sites that has been reshared to foreign governments, including adversaries.
In June, 2021, Senator Ron Wyden stated this:
“This information would be a goldmine for foreign intelligence services that could exploit it to inform and supercharge hacking, blackmail, and influence campaigns.”
He then sent out a letter to top ad networks and exchanges including Google, OpenX, Verizon, Twitter etc. to “name the foreign-headquartered or foreign-majority owned firms that they have provided bidstream data from users in the U.S. to in the past three years.” The list comprises 150 companies from China, United Arab Emirates (UAE), EU, Russia, Singapore among others.
What started out as a solution to define market efficiencies has become a purveyor of one of the largest global data collection schemes across a growing network of publishers that has enabled more invasive personal targeting of consumers without their knowledge or consent.
In turn, this has yielded many complaints since the arrival of the EU General Data Protection Regulation (GDPR) in 2018. Just a year later the GDPR received over 4000 complaints, many aimed at the tech giants, Facebook, Twitter, Instagram, WhatsApp, Apple and Google with the dominant complaints being consent, right of erasure, access rights and unfair processing of data. In 2018 and 2019, Facebook, LinkedIn, and Twitter had received complaints for unlawful processing of personal data for targeted advertising and behavioural analysis on their platforms.
We are now producing mechanisms to not only increase consumer understanding, but also to manipulate.
In October, 2018, the IAB Europe introduced the Transparency and Consent Framework (TCF) requesting users accept/reject ad trackers to help publishers comply with the new data protection rules. In the fall of 2020, from an investigation by the Belgian Data Protection Agency, they concluded the TCF did not comply “with GDPR principles of transparency, fairness and accountability, and also the lawfulness of processing”. What was also uncovered, with the help of Johnny Ryan, Senior Fellow of the Irish Council for Civil Liberties (ICCL), who filed the original complaints about RTB were these disturbing facts:
- Google’s RTB system sends data to 968 companies;
- a data broker used RTB data to profile people influenced the 2019 Polish Parliamentary Election by targeting LGBTQ+ people;
- a data broker built a profile with RTB data that allowed users of Google’s system to target 1,200 people in Ireland profiled in a “Substance abuse” category, with other health condition profiles offered by the same data broker available via Google reported to include “Diabetes,” “Chronic Pain” and “Sleep Disorders”
- IAB’s RTB system allowed users to target 1,300 people in Ireland profiled in an “AIDS & HIV” category, based on a data broker profile built with RTB data, while other categories from the same data broker include “Incest & Abuse Support,” “Brain Tumor,” “Incontinence” and “Depression”;
- A data broker that gathered RTB data tracked the movements of people in Italy to see if they observed the COVID-19 lockdown;
- A data broker that illicitly profiled Black Lives Matters protesters in the U.S. had been allowed to gather RTB data about Europeans;
- The industry template for profiles includes intimate personal characteristics such as “Infertility,” “STD” and “Conservative” politics;
As of June 2021, after years of RTB complaint filings to European Data Protection agencies that had yet to yield effective resolution, Ryan and the ICCL took the IAB Tech Labs, a standards body for the digital advertising industry based in NYC, to court in Germany, targeting ad tech who have profited from user profiling including Google, Amazon, Twitter, Facebook, Verizon as well as smaller lesser-known players including data brokers and data processors, handling user data within the RTB network. What Ryan argued in his filings was that the RTB systems were unable to properly ensure appropriate data security nor protection against unauthorized use, loss or processing, because of how they function.
Michael Veal is a tech policy researcher at the University of London and he dubs IAB’s TCF as “an attempt to legitimize an already flawed system”. He goes on to say, “The framework completely ignores the principle notion of data protection...Data needs to be kept secure, minimized and it can’t be disseminated among numerous parties – and that’s regardless of any transparency.”
The case is still pending, however, the outcome may create a significant precedent to hasten the resolution of the GDPR complaint backlogs.
The recent Senate Hearings on Protecting Kids Online and Facebook Whistleblower, Frances Haugen’s testimony uncovered disturbing insights about how the company’s own research confirmed Instagram’s effect on teenage girls, claiming, “2018 algorithm change which prioritized posts with high user engagement. It turns out lies and anger rank off the charts.” As per Haugen, “Facebook has realized that if they change the algorithm to be safer, people will spend less time on the site, they'll click on less ads, they'll make less money”.
The data garnered from Facebook, largely self-reported and volunteered by their user base, contributed to the effectiveness of Cambridge Analytica. The 2016 elections were manipulated by false news articles posted to organic timelines and amplified by bot accounts. Coupled with AI and hyper personalization, you have a recipe to not only predict future behavior, but also drive it’s direction.
What began under the guise of a “Netflix Prize” in 2006 to develop AI that would successfully predict user rating of movies for the purpose of future personalized film recommendations, has evolved to AI that is increasingly capable of performing personality analysis on-demand. Those with the biggest budgets compete in RTB on closed platforms—such as Google, Facebook and Twitter—to effectively control exposure to products, media and ideas or generally, “idea exposure”—a danger that Stanford’s One Hundred Year Study on AI (AI100) specifically called out.
As Dennis Wilson explains in his study, “The Ethics of Automated Behavioral Microtargeting,” by 2011, research showed that AI was capable of determining the political alignment of individuals based on their Twitter data, and by 2013, research showed that large amounts of data could be harvested from Facebook with high accuracy rates of AI’s ability to discriminate between homosexual and heterosexual men in 88% of cases, between African Americans and Caucasian Americans in 95% of cases, and between Democrat and Republican in 85% of cases.
The performance levels these platforms are capable of returning come from their special sauce, which include the derivation of psychological and emotional factors by holding users captive on their platforms. Behavioral and micro-targeting without access to these psychological factors may be less effective on the long-tail – that is, all those thousands of micro and niche sites that have very little traffic by comparison, however the manipulation messaging coupled with the use of fake accounts will amplify the harm, regardless.
Numerous studies were published this year examinining the psychological factors using Privacy Calculus Theory. These studies aimed to measure the negative and positive effects on people’s willingness to disclose personal information. Ironically, many of the papers describea positive correlation between an increase in meaningful privacy legislation, organizational-level privacy self-regulation and the users’ access to privacy controls with their willingness to share personal information “safely.”
We are in a time of “Market Shaping” and Public Policy, along with Responsible Tech Innovation, will have Critical Roles to Play
Can the RTB industry grow without data? Johnny Ryan has suggested that removing personal data from RTB system will solve the problem. CEO of AppNexus, an ad exchange begs to differ: “RTB without targeting, why bother!.....Making RTB untargetable kills it, and it’s not the right conversation to be having, it’s not an honest conversation.”
However, arguments have been made to take a step back and offer more contextual targeting instead. This alternative is basically what we had been doing in print magazines, cable TV shows, and radio pre-digital. That worked for decades and now we are coming back, full circle to contextual ads. Sadly, advertisers have wasted 10 yrs and a trillion dollars on digital ads thinking they were doing something evolutionary.
As long as there is demand for RTB the industry will grow. However, the legal efforts to contain it and improve its systems to limit the risks to individual data privacy and security, have seen few resolutions thus far.
Ad tech industry doesn’t view the resulting harms as compelling enough to derail their current ad models, as is clearly the case for Facebook. However, Haugen’s congressional testimony seemed to produce a turning point that will bring more accountability to Big Tech. Government, however, as one of the gravest offenders of acquiring and using people’s personal information in the name of national security, needs to reconcile those decisions with the stateside legislation on evolving data privacy measures.
So, it will come down to choice. The choice is for advertisers how to spend their dollars more responsibly, understanding the impact on their own bottom line and on a deeper societal level. What will spur advertisers to demand responsible advertising choices? Will legislation create precedence towards necessary change? The economics of supply and demand will determine whether RTB is sustainable, and whether the ad industry is due for a reckoning.
Post a Comment