It’s time to talk about the big controversy that’s got everybody buzzing online. That’s right, I’m talking about the viral fervor over Facebook CPM rates in the 2016 presidential election, baby! [loud klaxon noise, followed by a steam-train whistle, followed by tons of fireworks]
Over the weekend, former Facebook product manager Antonio García Martínez wrote a for-dummies explainer on wired.com about how Facebook’s automated ad auctions (in which advertisers bid for Facebook ad space in front of specific groups of users) work. Here’s the relevant bit:
Facebook uses a complex model that considers both the dollar value of each bid as well as how good a piece of clickbait (or view-bait, or comment-bait) the corresponding ad is. If Facebook’s model thinks your ad is 10 times more likely to engage a user than another company’s ad, then your effective bid at auction is considered 10 times higher than a company willing to pay the same dollar amount. […]
During the run-up to the election, the Trump and Clinton campaigns bid ruthlessly for the same online real estate in front of the same swing-state voters. But because Trump used provocative content to stoke social media buzz, and he was better able to drive likes, comments, and shares than Clinton, his bids received a boost from Facebook’s click model, effectively winning him more media for less money. In essence, Clinton was paying Manhattan prices for the square footage on your smartphone’s screen, while Trump was paying Detroit prices. Facebook users in swing states who felt Trump had taken over their news feeds may not have been hallucinating.
This got a lot of attention, because, well, the takeaway is that Facebook’s automated ad-buying platform is engineered to favor (outrageous, sensationalist, polarizing) ads like those shared by the Trump campaign, rather than those (vague, anodyne, platitudinal) ads posted by the Clinton campaign. Think “Crooked Hillary Wants to Put Criminal Aliens in Your House” versus “America Is Already Great.” While the broad mechanisms of Facebook advertising have been public since the beginning, people who aren’t in the advertising business haven’t always been paying close attention. Until now.
García Martínez’s article got an extra bump thanks to an endorsement from Brad Parscale, who ran Trump’s Facebook-ad-buying operation in 2016 and is set to run Trump’s entire 2020 campaign. Dollar for dollar, he said, the Trump campaign reached many more people than Clinton.
Jennifer Palmieri, the Clinton communications director, weighed in succinctly:
To explain what they’re talking about: Showing an ad to a user is an “impression.” The amount a publisher charges (or an advertiser pays) for 1,000 impressions is known as “CPM” or “cost per mille.” It’s one of the standard ways of measuring advertising costs online.
So did Facebook actually charge the Trump campaign less? There are two things worth noting here: The first is that Parscale and Palmieri probably don’t know what each other was being charged on Facebook — and, further, Parscale and Palmieri both benefit from the narrative that Trump was undercharged and Clinton over-. (Parscale can claim to be a digital-campaign Facebook genius, and Palmieri can claim that the Clinton campaign was screwed over by Facebook.)
The second thing worth noting that while García Martínez was asserting that Trump got a bigger bang for his buck on Facebook because his ads effectively mimicked the sorts of “provocative” content that perform well on Facebook (I’m trying not to write “clickbait,” but it’s the same genus), he didn’t himself claim, as Parscale and Palmieri did, that Trump’s CPM was lower.
Finally, there’s a complicating factor here, which is that, in addition to the viewers you’re paying Facebook to reach (known as “paid reach”), users might also share the ads on their own to other users (“organic reach,” for which you pay nothing). On Facebook, you could talk about your actual CPM — the amount Facebook charges you to put an ad in front of 1,000 people — and your effective CPM — the amount you’re paying to actually reach 1,000 people. Effective CPM is necessarily lower: If an ad you paid to place in front of 1,000 users for $10 is shared with another thousand users, you’ve actually reached 2,000 users for that $10. Your effective CPM is $5.
If, somehow, this blog post has not put you into an indefinite coma (CPM rates! programmatic ad auctions! the 2016 election!) then you’ll be happy to know the situation got even muddier yesterday afternoon, when Facebook executive Andrew Bosworth shared this chart, which — he says — shows Clinton’s CPM rate was lower than Trump’s.
“How???” asked all three people reading this blog post. “How could the rate be lower?”
Well. One problem here is that everyone is being a bit cagey about exactly what they’re talking about. Effective CPMs or actual CPMs? Paid reach or organic reach? Bloomberg’s Sarah Frier confirmed that Bosworth’s chart referred to paid reach only — is that what Parscale and Pamieri were referring to as well? And, what if I told you [jazz hands] there are two types of ad campaigns? In a thread on Twitter yesterday, García Martínez explained the difference between a direct-response campaign and a brand campaign.
Direct-response tries to get the recipient to take action by targeting a custom audience (buy a product, or, in Trump’s case, donate), which Facebook can do with pinpoint accuracy. A brand campaign, such as the type Clinton generally ran, is broader. Direct-response ads, because of their more precise targeting, usually have a higher CPM, so comparing CPMs for a direct-response campaign against CPMs for a brand campaign doesn’t necessarily tell us much about which campaign was favored by Facebook’s auction structure.
All of which leaves us more or less back where we started: Baffled. (To muddy it further: A couple of days ago, the Verge reported — via a single source, so take it with a grain of salt — “that Trump’s effective CPM averaged $0.06, compared with $1.06 for Clinton.”) What we have, ultimately, is a situation in which the metrics can be stretched and twisted and presented however you’d like. Obviously, in some sense, based on some comparison, among some demographics, Bosworth has shown that Facebook didn’t charge Clinton more. But we don’t know exactly what that comparison was, and whether or not it contradicts Palmieri, Parscale, or Garcia-Martinez’s stories.
Maybe Clinton was better at buying ads, because she had a lower CPM, or maybe Trump was better at buying ads because his effective CPM was lower and also he’s now the president. Maybe Facebook isn’t at fault and the users sharing the ads are responsible; maybe Facebook is ultimately responsible for creating a media ecosystem that allows someone like Trump to suck up attention. Maybe it’s cable news’ fault. Maybe nobody is at fault and that’s just how democracy works.
One thing to note, though, is that Facebook has expressed a willingness to release more complete data so long as the Trump and Clinton campaigns agree to that. This seems like a good idea. Or, at least, a better one than having this discussion via cryptic tweets and vague charts.