In a recent blog post, Facebook's vice president for ads, Rob Goldman, argues his platform's users aren't its product. Even though Facebook primarily makes money by selling targeted ads based on what it knows about you, Goldman says that the real product is the ability to connect people—ads merely exist to "fund that experience."

To help support that stance, Goldman paints a simple picture of the role advertising plays on Facebook, downplaying the information it collects about you. Using a hypothetical example about a small bike shop in Atlanta, he emphasizes that targeted ads help small businesses reach customers—like, say, female cyclists who live nearby—more efficiently. The example, which the social network also uses on a general page explaining how ads work, does represent how some companies use Facebook. But ad industry experts say Goldman's explanation leaves out many important realities of Facebook's advertising machine.

Relevant to What?

Four times in his blog post, Goldman stresses that Facebook's targeting mechanisms allow users to see relevant ads. But nowhere does he define what "relevant" means in this context. In some ways, it's broadly intuitive across all industries; advertisements for dentures or funeral insurance don't run on Nickelodeon for a reason. But beyond simple demographics, a "relevant" ad to a marketer might target a specific personality type, or perceived emotional state. It might also be designed to take advantage of an already vulnerable population. That can quickly get a lot more involved than just people who like bikes.

'There’s 60,000 channels and weird ways to combine them.'

Kane Jamison, Content Harmony

“We already have been seeing the results of negative segmentation we saw in the past before, like when cigarette companies were targeting low-income people,” says Juan Mundel, a professor at DePaul University who has studied Facebook advertising. Because the social network has so much data, it's possible to target hyper-specific audiences with extreme precision. That means, as Bloomberg reported in March, predatory advertisers can exploit Facebook's tools to sell shady products to the masses, like diet pills.

“Facebook also knows when you’re motivated to do something, when you’re feeling down, when you’re feeling all sorts of emotions,” says Mundel. The social network leverages that information for advertisers; the Intercept discovered earlier this month that the company has developed a new service designed to predict how consumers will behave in the future, like when they're likely to switch from one product brand to another. That level of psychological parsing goes far beyond what Goldman outlines.

Goldman is right to point out that Facebook has much in common with traditional forms of advertising like television and print, but the difference is companies who use Facebook have a near-endless number of data points with which to target their ads, and can show them to much narrower slices of the population. "Facebook is the same thing, but there’s 60,000 channels and weird ways to combine them as well,” says Kane Jamison, the founder and managing director of Content Harmony, a marketing agency that frequently uses Facebook to advertise.

What Facebook Knows

Throughout Goldman's post, he stresses that users can control their ad experience by visiting their Ad Preferences menu. At the top of the screen, you will see Your interests, which Facebook says it generates based on your activities on Facebook, such as pages you may have liked. It's not clear whether some of these categories are algorithmically generated, and Facebook would only say they are based on past actions on the platform.

The categories range from intuitive to bizarre. Mine, for example, include head-scratching topics like "Laser," "Steel," "Everything," and "Authority." If you hover your cursor over each one, Facebook ostensibly tells you why it first appeared: "You have this preference because you clicked on an ad related to Everything." Huh? Advertisers have a stunning number of categories to sift through—ProPublica has collected over 50,000, including those only marketers can see. You can remove any interest associated with your profile by clicking the X in the top right-hand corner.

Toward the bottom of the screen, under Your information > Your categories, you can see other information Facebook has collected about you, based on your activities both on and off the platform. There, you might notice that Facebook has categorized you as a "newlywed," "away from family," or "close friends with expats." You will likely also see the type of browser and phone you use. You can remove any of these categories as well.

The Ad Preferences menu is hard to find, difficult for the average user to parse, and unclear about how Facebook compiled the information. Goldman is right that the menu allows you to tailor your experience—you can even opt-out of ads for alcohol, children, and pets entirely—but it's not designed in a way that allows users to understand how their information is being collected in the first place.

"They only show you four rows of content. The rest you have to open up with a pull-down menu, it’s designed in a way that’s not user-friendly,” says DePaul's Mundel.

Mix And Match

All the ways marketers and advertisers can combine these data points to make inferences about users also remains unclear. Of course, they can be used in fairly standard, innocuous ways. A local coffee shop, for instance, might understandably target people who already like Starbucks and pages like "Need My Morning Coffee." But a study published in Proceedings of the National Academy of Sciences late last year found it was effective for marketers to also target specific personality types, deduced by a single Facebook page users had liked. For example, the researchers assumed people who liked the page "Computers" were more introverted, and targeted ads to them based on that personality type. They found users were significantly more likely to click on the psychologically tailored ads.

'It’s designed in a way that’s not user-friendly.'

Juan Mundel, DePaul University

Facebook marketers rarely however stick to one category to target you. To identify an audience who will engage with a specific advertisement, marketers need to know how to best combine these categories. “Interest doesn’t mean you like the topic, it just means you keep looking at it or something like that,” says Content Harmony's Jamison. A user might, for example, show an interest in Donald Trump, but it doesn’t mean they like the president, it just means they read stories related to him. To target people who genuinely like Trump, marketers would need to combine an interest in him with another category. But users don't always know how these combinations happen, or why, making it difficult to know the exact reasons you may be seeing an ad targeted to you. Facebook does allow you to click the three dots in the corner of any advertisement and view some information about why you're seeing it, but it's difficult to get a clear picture.

"Whatever information goes into sense-making about an individual and grouping people into different groups and segments is much more than demographic," says Saleem Alhabash, a professor at Michigan State University and the co-director of its Media and Advertising Psychology Lab. "It's the things you interact with, every URL you click on—not only on Facebook but elsewhere on the internet—also how you interact with your friends, all of this gets muddled into making sense of you and providing you with relevant ads."

In his blog post, Goldman says that advertisers on Facebook use three kinds of information to target you: data you've provided to the social network (like your age, or a page you've liked), information an advertiser may already have about you (like your email address), and your activity off Facebook (like your browsing history). “The fourth one that is very clearly missing is information from third-party data brokers,” says Jamison.

In March, Facebook announced it would wind down its Partner Categories program, which allows advertisers to use data from third-party data brokers like Axiom to target campaigns. It's what allows marketers to target you based on estimated income, for example. In exchange for use of their data, Facebook gives the data broker a cut of the advertising sale.

Partner Categories officially shuts down next month in the European Union—the same day a new, strict privacy law, the General Data Protection Regulation, goes into effect. Elsewhere, Partner Categories will be winding down over the next several months. In other words, it's still in effect—but Facebook makes no mention of it in its explanation of how advertisers target users.

Over the last several months, Facebook has taken steps to make its advertising platform more secure and transparent. In October, the social network announced it was hiring 1,000 more people for its global ads team over the next year, and said it would start making all ads from a company visible to anyone, not just the ads targeted at them. In April, Facebook also imposed greater restrictions on organizations who buy ads for hot-button issues like immigration.

These are good steps, and will likely make Facebook more resilient against the kind of misinformation campaigns that exploited its tools during the 2016 presidential election. But for the average user, Facebook's advertising system is still a black box. By under-explaining its complexities, Facebook likely also only makes the confusion worse.

Facebook Dodges

Read more: