We asked three media agencies to come up with media plans and budgets for someone looking to swing an election in Michigan and Wisconsin using facebook.
Facebook is currently embroiled in an investigation regarding the spread of viral fake news and ads by Russian groups leading up to the 2016 presidential election. The company is expected to testify before two congressional committees on Capitol Hill in November.
Facebook, which declined to comment for this story, has said it discovered roughly $100,000 in ad buys between June 2015 and May 2017 associated with roughly 3,000 ads. CNN reported that a number of these ads specifically targeted Michigan and Wisconsin, two states where US President Donald Trump won by approximately 10,700 votes and 22,700 votes, respectively.
The money spent may seem insignificant at first glance, but Facebook's advanced and granular targeting options may have given that dollar amount significant reach and engagement. Add to that the fact that operatives — whether it's a candidate, a PAC, a non-profit or even a fake account run by groups linked to Russia — are not required to disclose spending money on online ads, and it becomes a bigger concern.
To see how easy it would be for just about anyone to use Facebook's targeting parameters to reach their desired audience, we asked three media agencies to come up with media plans and budgets — in this case specifically for someone looking to swing an election in Michigan and Wisconsin.
It is worth noting that buying and distributing ads is not nefarious in and of itself. Companies, advocacy groups, political groups, politicians, and others all use Facebook's platform to try to sell you their stuff or get their messages across.
That's how these media agencies were able to model out the scenarios below — they help advertisers do this
The analyses of course make certain assumptions: that a campaign can target swing voters; that the ads actually influence the number of people who are projected to be influenced; that the people who are targeted haven't already made up their minds and are open to being influenced; that people see these ads at a relevant time (i.e. before an election).
And finally, Facebook reportedly will soon start tightening the screws on political advertisers using its platform, manually reviewing ads that are targeted to people based on "politics, religion, ethnicity or social issues." In other words, some of these projections could work out differently in the future.
According to Ben Kunz, EVP of marketing and content at Mediassociates, a political operator looking to swing an election in Michigan and Wisconsin would need to pinpoint and target undecided voters using various data tactics. After that, "it wouldn't cost much in ad spending to sway their opinions," he said.
Mediassociates built a model based on a basic rule of thumb of digital advertising, which is that 1 out of 2,000 people (or 0.05%) who view an ad will respond or take action on the message. Their model works backwards from the exact numbers of people by which Trump won both Wisconsin and Michigan, although, in reality, it's possible that someone targeting swing voters could try to reach more people than that and therefore might spend more money on their campaign.
Donald Trump won Michigan by 10,704 votes. So assuming that only 0.05% viewers will react to an ad, a person would need to aim for about 21.4 million advertising impressions. Given that Facebook ads cost about $2.00 for every 1,000 impressions, Kunz's team estimates that to sway about 10,700 voters you'd need a budget of $42,800.
"People like to think they can't be persuaded, but the math says it works."
As for Wisconsin, Trump won the state by 22,748 votes. Using the same model described for Michigan, that would suggest a budget of $90,992.
Putting those two costs together gives you about $133,792, which works out to about $4 per voter. We included their table of the assumptions and calculations below.
"In reality, targeting this exact 'swing voter' population might take more effort, since some of your ads will reach the wrong people ... and competitors might be fighting you with similar tactics," Kunz told Business Insider. "But a clever political operator would just spend a little more. And for a few hundred grand, he or she could tip an entire presidential election."
"That's the scary power of Facebook. Its targeting is really, really powerful," he added. "People like to think they can't be persuaded, but the math says it works."
As an aside, Kunz also pointed to an interesting feature on Facebook's system, which is that you can actually target people by using gerrymandered Senate and House district borders by punching in individual zip codes. Again, this is not nefarious in and of itself, but does demonstrate the power of Facebook's ad targeting system, and could be of relevance given that the Supreme Court is currently looking at a case on gerrymandering.
Facebook Ads Manager has both broad targeting capabilities and very specific targeting capabilities. An example of the latter would be political affiliations broken down by liberal, conservative, and moderate leanings.
Michael Dobson, group director of social media at Crossmedia, looked at four targeting approaches to see the possible reach per audience within Wisconsin and Michigan using such affiliations.
For his first three approaches, he looked at liberal, conservative, and moderate target audiences. A campaign running for six weeks that reached a given user two times every seven days would be able to reach approximately 74% of users who fall under those three audiences, he said.
His total estimated cost for such a campaign was $250,000. Breaking it down, $100,000 was dedicated to "liberal and very liberal" audiences, $100,000 was dedicated to "conservative and very conservative" audiences, and $50,000 to "moderate" audiences.
Dobson also came up with a fourth approach going in a different direction. Instead of targeting the political leanings of audiences, he selected a "sensitivity issue," or a controversial topic that attracts a lot of tension, like gun control.
Over the same six-week time period, he says such a campaign could reach approximately 66% of the over 3 million Facebook users in Michigan and Wisconsin with a budget of about $50,000.
Such a strategy would target people who Facebook has identified as those who are interested in the topic based on their "likes," what they search, what they read online, etc. The budget was smaller than the budgets for targeting "liberal and very liberal" and "conservative and very conservative" audiences because the audience is smaller, he said.
Essence went a slightly different route. The team put together three highly specific targeting "buckets," which they created based on researching the demographics and interests of the undecided and moderate voters, and by using Facebook's targeting data:
They also looked at traditional swing districts in the two states (Wisconsin's 1st and 3rd districts; and Michigan's 5th, 6th, 9th, and 11th districts).
For reaching 60% of the targeted audience, they put together two different scenarios based on how many times an ad would reach a given person (two times every seven days versus three times every seven days), and came up with the two budgets: $188,700 and $283,050.
The above three examples are more generally about targeting ads at voters for political purposes in Michigan and Wisconsin. Similar strategies can be applied to more specific cases such as targeting ads for fake news content.
Still, there would be some differences. In Essence's above strategy, the team said one of its target groups would include college graduates. Jeff Rayvid, a data analyst at the firm, however, told us that if he were to create a strategy for targeting ads for fake news specifically, he probably wouldn't target college-educated users.
Larry Kim, CEO of MobileMonkey actually ran an experiment where he created a fake news website, created a Facebook page for that website, and then created an ad to promote that page, which he described in detail in a post originally written on Medium.
He says his ad was approved "within minutes" and reached 4,645 people, generating 44 "likes," 27 shares, 20 comments, 3-page likes, and approximately 200 website clicks. He paid a grand total of $53.58, noting that it's "remarkably cheap" to push fake news stories.
"Facebook claims that their ads can have a profound impact on a user’s searching and purchasing behavior. It’s not hard to believe that it could impact voting behavior and even an election outcome in battleground states where the margin of victory was just a few thousand votes," Kim wrote.
"Facebook ads were and remains to this day, a highly effective vehicle for the distribution of fake news intended to alter public opinion both in the USA and in other countries worldwide where Facebook advertising costs are substantially lower."