Data Minimization: Regulating the Ineffective, Irrelevant, and Invasive Practice of Surveillance Advertising
August 8, 2023 |
This is the fifth blog post in EPIC’s series on Data Minimization. We have previously discussed data minimization as a framework to curb harmful commercial surveillance practices, harms that stem from out of context secondary data uses, data minimization as a pillar of data security, and data minimization as a tool to protect health data privacy. This post discusses surveillance advertising and questions whether this invasive business model provides consumers with any significant benefits. Bottom line: surveillance advertising does not provide consumers with more relevant, higher quality, or lower priced ads, meaning the harms associated from this model are not outweighed by any benefits to consumers. (Note that this post does not discuss the harms from surveillance advertising in detail, which EPIC has written about extensively.)
The FTC’s Mandate to Determine Benefits to Consumers
As we’ve said before, the Commission should use its proposed commercial surveillance and data security rulemaking to promulgate a data minimization standard that will limit harmful data practices like surveillance advertising. This post analyzes whether the claims made about surveillance advertising by industry actors accurately reflect their purported benefits to consumers.
If and when the Commission promulgates a proposed rule, the Commission will have to determine that the commercial surveillance practices targeted by that rule are either deceptive or unfair. To determine that a practice is unfair, the Commission will have to find that it (1) causes or is likely to cause substantial injury (2) that is not reasonably avoidable and (3) is not outweighed by countervailing benefits to consumers or competition. This post addresses part of the last prong of the test—namely, the fact that surveillance advertising provides no benefits to consumers that outweigh its harms. We urge the Commission to use its resources to verify claims made about the effectiveness of targeted ads.
Many large tech companies claim that surveillance advertising provides consumers will well-tailored, relevant ads that consumers want to see as they browse the internet. The theory goes that gathering vast amounts of personal information about an individual will enable publishers and retailers to personalize their internet experience with ads for products and services they want to purchase. In reality, most ads that an individual sees are not relevant to them and most of the ads are for lower quality products and services. There is simply no worthwhile tradeoff for the invasive, harmful nature of surveillance advertising for consumers.
As we have previously explained in our comments to the Commission, the profiling and targeting of consumers online is an unavoidable, injurious practice. A 2021 FTC report found that a single ISP had 370 million consumer relationships and that another ISP served one trillion ad requests monthly. (These figures are likely higher today as the online advertising industry has continued to grow since 2021.)
The average consumer spends 7 hours a day online. This means we see hundreds or even thousands of ads each day. When browsing the internet, reading emails, or online shopping, we are bombarded with digital ads. They can be annoying cookie banners that take up extensive screen space, brightly colored pop ups, or images that distract us as they change every 15-20 seconds. Not only are digital ads annoying, but they are becoming more numerous: we are seeing more ads online than ever. In fall 2022, 52% of Facebook users and roughly half of YouTube and Instagram users reported seeing more ads on their respective social networks. Because we constantly see ads online, it is even more annoying that those ads are largely irrelevant, repetitive, and low quality.
Targeted Ads Are Not Effective
Targeted Ads Show Lower Quality Products with Higher Prices
It can be difficult to measure the effectiveness of any digital advertisement for several reasons. A person may see an ad online for yogurt and then be more likely to purchase that brand of yogurt next time they are in the grocery store. A public health campaign may consider its marketing strategy success if many people see its banner to promote a vaccine, regardless of how many clicks the banner gets. Perhaps a user clicks on an ad for shoes from a department store, then searches for the shoes and ultimately purchases the shoes at a lower price from the shoe store itself. Consider also a person who does click on an ad for shoes and purchases them—but then continues to see the same ad for the same shoes pop up everywhere for weeks. Perhaps a person who lives in Florida searches for a winter coat for a family member as a gift but ultimately chooses something else and continues to see ads for winter gear that are irrelevant to them. Does a person clicking on an ad mean the ad was effective? What about a purchase that is sparked by the ad but not ultimately traced back to the ad? On top varying measures of efficacy, many large publishers do not reveal who advertises on their sites. It can be difficult for any one person to know whether the ads displayed on their own browser are improving their experience as a consumer online.
Three economists recently published an investigation into behavioral advertising and consumer welfare that found that targeted advertisements did not show users highly relevant ads from known vendors at lower prices The authors, Eduardo Abraham Schnadower Mustri, Idris Adjerid, and EPIC advisory board member Alessandro Acquisti explained that the value consumers derive from targeted advertising is largely posited rather than empirically demonstrated. In other words, large publishers that display ads and advertisers often claim that surveillance advertising provides great benefits to consumers, but they never prove this. In fact, the authors’ research found the opposite. They examined about 1,000 participants and online ads that they saw across two studies. The experiments asked the participants to browse randomly selected websites on their own computers and share the URLs of the ads with the researchers. The researchers then used automated scripts and Google to search for competing products with those advertised to the participants. Then, participants were shown the products that had been displayed to them (ad condition), competing products obtained from the organic results of online searches (search condition), and randomly selected products (random condition). They recorded objective measures like price and vendor quality and self-reported metrics such as relevance and purchase intentions (the self-reported likelihood of a participant to purchase the product).
First, the authors found that most targeted ads for products and search results for competing products were dominated by a minority of vendors (sellers)—yet targeted ads were more likely to present participants with lesser-known vendors than products from a search. In other words, when a person sees a targeted ad for a product while browsing the internet, the limited number of vendors represented in the targeted ads they see will tend to be less popular than the ones found through a search. Next, the authors found that purchase intentions are higher in targeted ads displayed on websites and searches than in a random condition due to higher product relevance in the ads and search conditions. However, product relevance is low in absolute terms even in the targeted ads displayed on websites. Finally, the authors found that targeted ads displayed on websites are more likely to be associated with lower-quality vendors and higher prices for identical products compared to competing alternatives in search results. A consumer is better off—i.e., will tend to find higher quality and less expensive products—by searching for a product than by clicking on a targeted ad on an unrelated webpage. The study also found that higher purchase intentions and relevance in targeted ads on websites were driven by a participant previously having searched for the advertised product.
This study is critical to understanding whether targeted advertising provides consumers with benefits. It provides critical evidence that, no, consumers do not enjoy a better consumer experience through surveillance advertising. In fact, a consumer will see higher-quality and less expensive products when they search for the product itself. In the cases where targeted advertising is more relevant to the consumer and the consumer has a higher purchase intention for the product (the second part of the authors’ investigation), it is typically due to the fact that consumers already searched for the product and are now seeing ads for it. Of course a consumer is more interested in an ad for a product that they have previously searched for. Consumer searches and research into products will provide a better experience for consumers than targeted advertisements.
Targeted Ads are Largely Irrelevant
Targeted advertising uses parameters such as gender, location, and age to deliver ads to individuals and groups. Most of the parameters are derived from profiling users’ behaviors because users do not generally directly provide adtech companies with their gender, age, address, and so forth. One study researched the targeting parameters purchased from adtech data brokers and found that a single ad parameter—gender—was only accurate 42% of the time. In other words, a company that attempted to target ads based on gender was less accurate (42%) than a random control (50%). The study explained: “Gender accuracy ranges from 25.7% to 62.7%, with an overall average of 42.3%. Given the benchmark for correct gender classification is about 50%, or the natural distribution of gender diversity in the population, using data brokers to assess online browsing profiles for gender appears on average less efficient than using nothing.” With two target parameters, gender and age, the accuracy was only 24% on average. Insider Intelligence reports that Facebook’s system of inferring user interests yielded incorrect results about 29% of the time. Adalytics’ Krzysztof Franaszek, who tracked the digital ads served to him for a month, determined that 74% of those ads were not relevant to him and that nearly 40% were from a single company.
Recent reporting confirms this. In May, the New York Times exploredreported on the increasingly annoying, irrelevant, and repetitive ads that we are bombarded with online. The paper noted that “ads that few people want to see suddenly seem to be everywhere” and that fraudulent ads have popped up more often on Twitter. On Instagram, “bizarre” ads for products from obscure merchants are promoted by Amazon. YouTube has been infiltrated with ads that impersonate popular creators in order to scam viewers. In order to continue to generate revenue wherever possible, large companies have sold more low-quality ads.
Publishers Use Targeted Advertising to Deceive Advertisers
Despite large publisher’s promises to advertisers that their ads will be seen by their targeted audiences, large publishers fail to prove their claims. Bots can inflate ad metrics by clicking on ads to trick marketers into thinking they are more effective than they are. Indeed, 15% of programmatic ad spending disappears somewhere in the supply chain, and about half the money spent by brands goes to digital middlemen. This system is not effective for brands who have little control or ability to verify the efficacy of their digital ads. If a user leaves a website open but is not in view, that unseen page may continue to unload and load new ads from a network every 15-20 seconds. This practice also inflates ad metrics as publishers appear to display dozens or hundreds of ads (depending on how long the page stays open in the background) without a user seeing the ad at all. This practice makes excessive use of individuals’ network data and can slow down their browsing.
This isn’t a new phenomenon. In 2018, a class action accused Facebook (now Meta) of inflating its metrics for “Potential Reach” of an ad, a number that indicates how many people might see an advertisement. This led advertisers to spend more money on the platform in hopes of meeting that inflated reach metric. According to leaked documents, some of Facebook’s top executives—including COO Cheryl Sandberg—were aware of the exaggerated reach potential. Perhaps this is unsurprising because: Meta’s personal data practices have caused serious privacy harms for nearly two decades, and up to 98% of its revenue come from advertising. A Facebook product manager apparently responded to executives’ concerns that changing the metrics to become more accurate would significantly impact revenue by saying “it’s revenue we should have never made given the fact it’s based on wrong data[.]”
Does surveillance advertising provide consumers any benefits that outweigh its significant harms? No! Targeted ads are only more relevant and have a higher purchase intention when a consumer has already searched for the product. Large tech publishers inflate their ad metrics so that advertisers will pay more for ad displays. Ads that are served on webpages not in view also inflate ad metrics and waste users’ network data. Fraud and bots bombard consumers with low-quality ads and can mislead advertisers as to how many views an ad receives.
EPIC has written previously about the serious injuries that stem the harmful practice of surveillance advertising. For years, large tech companies claimed that the benefits of this business model outweighed those harms. Yet as we have seen, the online advertising ecosystem incentivizes excessive surveillance of consumers but does not even produce more effective ads than a basic search would. Surveillance advertising relies on firms that collect vast amounts of personal information about us, store that information indefinitely, and link that information with our devices so that our profiles follow us wherever we go online. This invasive surveillance system ultimately serves no better ads to consumers.
The Commission should use its commercial surveillance rulemaking to finally regulate this harmful advertising system. At the very least, the firms engaging in such surveillance should have to prove the purported benefits of these practices. For decades, firms have said they need nearly unfettered access to our personal information in order to make the consumer experience online better. So where’s the proof?