Facebook’s financials showed the company made $84.2 billion in advertising revenue in 2020
—its major source of revenue. Facebook sells itself as a great way to tailor and deliver ads to highly specific audiences, but how that ad-targeting process works is a mysterious combination of audience selection by advertisers and algorithmic decision-making from Facebook.
Facebook has often played a double game in how it describes its advertising technology, on the one hand downplaying its intrusiveness
in communications to users of the platform and on the other pitching companies on the ability to predict future user behaviors
and potentially alter those behaviors with advertising.
We can’t completely unravel how Facebook ad targeting works, but we can use our Citizen Browser panel to see if certain groups get more ads than others. Our panel included more than 2,500 people across the country from December 2020 through February 2021 who automatically shared their feed information with us. And as we sifted through the data generated by our panel, we turned up an interesting observation: The average number of ads seen varied by location and demographic characteristics, with state of residence accounting for the largest variation in ads seen.
Discovering Ad Hotspots
To lay the groundwork for our research into Facebook ads, Micha Gorelick
—a consultant data scientist for the Citizen Browser project—created a new metric to determine which groups of people were seeing ads more than others: ads per capture.
Ads per capture is a metric that accounts for the fact that the Citizen Browser database is compiled from a series of “data captures” taken from our panelists’ feeds multiple times a day. Each data capture contains a chunk of around 10 to 20 Facebook posts that appeared at the top of the news feed at that moment. (You can learn about the Citizen Browser panelists and data collection process in depth in our methodology
So ads per capture (APC) is a number that simply represents the average number of posts that are ads in each data capture we get. In other words, it’s a rough estimate of how many ads a user might see at the top of his or her news feed. And once we have that number for the total dataset, we can refine our understanding by breaking it out by different demographics.
For one thing, we find out that the largest variation in APC corresponds with panelist location. In data from panelists based in Georgia and Pennsylvania, we received an average of just under 2.4 APC between Dec. 1, 2020, and March 1, 2021, with Georgia marginally the higher of the two. These states plus Massachusetts, South Carolina, and Illinois were the top five locations ranked by APC. (As The Markup previously reported
, Georgia was the only state in which political Facebook ads were allowed from December to January, which may have impacted the results.)
At the lower end of the spectrum, data from panelists in Iowa, Connecticut, Louisiana, and Wisconsin contained around 1.9 APC, and data from North Carolina averaged 1.96 APC. We discarded data from geographic regions and other demographic groups that comprised less than one percent of our panel size.
Highest and lowest ranked states by APC