What We Don't Know About What Facebook Knows


If Facebook data has such a far reach, why does it contain basic errors of plausibility?

Facebook claims it can reach more young people in the United States with advertisements than are actually alive, according to a new report.

Last week, a trade magazine in Australia looked at the “reach” statistics that Facebook gave for that country, and found that the company estimated it could reach more young people than Australia’s census.

Now, the firm Pivotal Research has done the same thing for America. “Through Facebook’s Ads Manager we can see that Facebook claims a potential reach within the United States of 41 million 18- to 24-year-olds, 60 million 25- to 34-year-olds, and 61 million 35- to 49-year-olds,” they wrote. “By contrast, U.S. Census data indicates that last year there were a total of 31 million 18- to 24-year-olds, 45 million 25- to 34-year-olds, and 61 million 35- to 49-year-olds.”

That’s an overstatement of 10 million people among 18- to 24-year-olds, and 15 million among 25- to 34-year-olds.

A Facebook spokesperson told The Wall Street Journal that the reach estimates “are not designed to match population or census estimates. We are always working to improve our estimates.”

Which, sure. Except that you’d probably want your estimates to match reality as closely as possible, and showing 30 percent more people than could possibly exist seems like a problem.

The larger context is that Facebook has been eating ever larger chunks of the digital-advertising market, soaking up ad dollars in large part because of the scale of the company’s user base (a representative example of coverage: “Facebook’s Scale Casts a Shadow Over Publishers”).

If just some of that scale is not real, that appears significant, even if, as the Journalnotes, “the impact of the discrepancy isn’t clear” for advertisers—at least not yet. (Several advertising associations did not immediately respond to requests for comment.)

It’s easy to see how such estimates could go wrong. People create more than one account. They switch devices. They lie about their ages.

But that strikes at the heart of the other big lure of Facebook. This is the company that’s supposed to have the most and best user data. They’re supposed to have solved the problem of matching up real people with online personae. That was the point of a huge London Review of Books essay on the company. “Facebook, in fact, is the biggest surveillance-based enterprise in the history of mankind,” writes the journalist John Lanchester.

And yet, on Instagram, which shares data with Facebook and uses its parent company’s back-end advertising infrastructure, the company’s algorithms continue to believe that I live in Washington, D.C., and occasionally serve me local advertising to that effect.

This makes some sense, given that The Atlantic is headquartered in D.C. But I only lived there for 18 months, and I left in 2011.

Granted, I’m just one person, and anecdotes aren’t data. But I’ve been posting pictures from and receiving bills in Oakland for six years. Shouldn’t the “biggest surveillance-based enterprise in the history of mankind” know that?

Maybe these demographic notes matter far less than it’s thought for advertising—age, location, that kind of thing. Because the truth is, I’ve even purchased hundreds of dollars worth of sneakers and clothes through Instagram ads. They do understand what I want to buy, even if they don’t know where I live.