Introducing ‘Decoder’

12th Oct 2020 | Share:

Using our new ‘Decoder’ tool, we’ve grabbed and decoded every URL linked to in Trump and Biden Facebook ads since July 1st*. 

Each URL tells us a little about what the campaigns are trying to accomplish with the ad. By putting them all together, and looking at the data in aggregate, we can see deeper into their advertising efforts than ever before.

For example, the Trump campaign uses 23 variables to help them track the performance of their ads. These include the audience segments and states they want to reach, the targeting methods being used, how they’re optimizing the ad to be delivered to just the right types of people, right down to the width, height and length of the video clips they’re using.

(A Trump campaign URL – the interesting bit starts utm_campaign=…)

What does that mean? We think it’s best to explain using examples, so based on data from nearly 300,000 (and counting – data is added daily) Trump Facebook ads, here’s every time he:

  1. Uses Facebook’s Custom Audiences to target voters on his email list, have visited his website, who use his app or whose mobile number he has.
  2. Uses Facebook’s Lookalike Audience targeting to reach new supporters.
  3. Targets people on Facebook based on their behaviours (such as liking Sean Hannity, the NRA and Ted Nugent)
  4. Asks military veterans to buy campaign merch
  5. Runs a ‘persuasion’ ad.
  6. Runs an ad to acquire new supporters
  7. Targets people who might be interested in him, with ads optimized for engagement.
  8. Targets people who look like his supporters, with ads designed to get them to like, comment on or share them.
  9. Targets people who haven’t taken an action yet
  10. Targets non-donors donors
  11. Targets his monthly donors
  12. Targets just women (1.53% of the time)
  13. Tries to reach just men (0.88% of the time)
  14. Runs Instagram-only ads (just 0.2% of the time)
  15. Targets people in Florida
  16. …in Pennsylvania
  17. …in Ohio
  18. …in Michigan
  19. …in Wisconsin
  20. Tries to sell hats (and other stuff)
  21. Optimizes his ads to get more engagement
  22. Runs a video ad with a 60 second spot
  23. The same, but for a 30 second spot
  24. Targets voters over 55, telling them Biden is too old
  25. Targets older voters with ads about prescription drugs
  26. Accused Biden of wearing an earpiece to try and raise end-of-quarter money
  27. Accused the left of ‘outrageous accusations’ about Mount Rushmore
  28. Asks people to vote early
  29. Made a ‘hard ask’ of supporters, looking for money right now.
  30. Made a ‘soft ask’ of supporters, looking for money later.

And much, much more… what can you find? Let us know!

We’ve been fighting for transparency for political ads for the last three years and this type of work takes us a step closer. We’ll shortly publish this dataset in full, along with the methods we’ve used to create it, but the arguments we make persist:

  • This type of data helps people better understand how campaigns work, what they think of voters and how platforms ultimately work. More is better.
  • That this data is available is a massive step forward vs. 2016, but we’ve had to work harder than we’d like to get it, because it’s only available through taking additional steps versus what’s available directly from the Facebook Ad API.
  • While we welcome the data and many of the steps Facebook has taken since we started to call for greater political ad transparency, governments HAVE to step up and define the new rules democracy needs in the digital age.

Decoder was conceived and built in partnership with Russell Garner.

Please consider supporting our work by making a one-off or monthly donation.

Support our work by making a one-off or monthly donation

One-off donation Monthly donation

Download the Who Targets Me browser extension

Chrome Firefox Edge Safari