Facebook VP Andrew Bosworth shares chart that shows the Trump campaign paid slightly higher CPM prices compared to Clinton in 2016 election, disputing reports
in a way. See update. http://www.washingtonpost.com/ ... Sarah Frier / @sarahfrier : I'm told this data is only counting paid reach, not organic. So the ads could have spread further (and by that measure gotten cheaper) for either campaign based on how FB users reacted to them https://twitter.com/... Boz / @boztank : After some discussion we've decided to share the CPM comparison on Trump campaign ads vs. Clinton campaign ads. This chart shows that during general election period, Trump campaign paid slightly higher CPM prices on most days rather than lower as has been reported. http://twitter.com/... Jason Kint / @jason_kint : I find it crazy press is allowing Boz to be the spin for Facebook. Demand an actual spokesperson or Sandberg if you're going to allow them to refute something. How do you even hold him accountable? He's now in the AR department. http://twitter.com/... Antonio García Mtez / @antoniogm : National averages won't tell the tale on a very complicated story. This needs to be broken out my market, action type, etc. to make any sense at all. Boz / @boztank : Prices depend on factors like size of audience and campaign objective. These campaigns had different strategies. Given the recent discussion about pricing we're putting this out to clear up any confusion. See also Mediagazer
Context & Ripple Effects
This lands three days after Wired's account of how the Trump campaign increased its Facebook ad purchasing power with provocative content and tools like Custom Audiences and Lookalike Audiences increased its ad purchasing power — reporting that fed a narrative the campaign bought reach on the cheap. Bosworth's chart pushes back directly: during the general election, Trump paid slightly higher CPMs than Clinton on most days.
The rebuttal comes with a methodological asterisk Sarah Frier flags immediately: the data counts paid reach only, excluding organic spread, so either campaign's effective cost per impression could look different once user reactions amplify ads for free. That caveat matters because the auction prices ads partly on predicted engagement.
First-order effects
- Facebook is pulled into publicly litigating its own ad-auction data, with Bosworth's chart contradicting the cheaper-CPM narrative that grew out of coverage of the campaign's provocative-content playbook.
- The paid-reach-only scope of the chart leaves the core question open: if organic amplification made Trump's ads effectively cheaper per person reached, both sides of the dispute can claim the same dataset.
Second-order effects
- Pressure builds on Facebook to release fuller campaign spending and delivery data itself — which materializes weeks later when its internal white paper put Trump at $44M versus Clinton's $28M across 5.9M ad versions internal white paper — making the platform, not outside researchers, the arbiter of election-ad facts.
- Campaigns and their media buyers learn that CPM comparisons are only as good as their reach definitions, pushing future election post-mortems toward granular delivery data like the micro-targeting breakdowns Facebook later supplied for 2020 micro-targeting data.
Third-order effects
- Political ad pricing on social platforms hardens into a standing accountability battleground where the platform's own disclosures become the primary source — a structural dependency that recurs through 2019 data-harvesting investigations data-harvesting investigation and the 2020 cycle.
- If engagement-weighted auctions systematically reward provocative political content with cheaper effective distribution, the debate shifts from sticker-price CPMs to whether auction design itself needs disclosure rules — a question regulators would have to answer with platform-supplied data.
The trend: Platform ad-auction economics are becoming the central evidence base in election-integrity fights, forcing Facebook to publish its own campaign data rather than cede the narrative to outside researchers.