Targeted Digital Video Ads in State Legislative Races: A Case Study

Last fall, I was lucky enough to help out with a very cool voter-turnout program employing sophisticated targeting to help elect down-ballot Democrats in two battleground states. The tools? Digital videos created by The Hometown Project and tailored to specific targeted state legislative and congressional districts, placed in front of priority voters via Facebook and programmatic video ads. The secret ingredient? Actors, comedians, musicians and other “energizers” recruited for their local ties, speaking directly to voters in their hometowns or home regions. My favorite? Ernie Hudson, the fourth Ghostbuster, featured above!

The people behind The Hometown Project started building out the idea soon after the 2016 elections. Coming from the entertainment world, they realized that they could mine their own personal and industry connections to try to bring influential voices to political races that usually don’t draw star power. They worked out a process to identify celebrities who might volunteer to help in state legislative races back home, first in off-off-year elections in places like Virginia and later in the midterms. Once energizers signed up, the Hometown organizers then got them to record videos touting specific local candidates.

Before 2020, Hometown had mainly been forced to hand off the videos to the candidates themselves or to political organizations working in the state, though they’d experimented with digital advertising. They brought me in to kick things up a notch last year, and we were privileged to get to work directly with the Democratic state parties in Michigan and Ohio. With a real promotion budget in hand, we were ready to run digital ads to make sure that the right voters saw the videos, and often.

By the end, Hometown had created literally dozens of district-specific videos in those two states alone, and we ran roughly 4.6 million video ad impressions to promote them. 2020 was a tough year for down-ballot Democrats, but several of our candidates won by narrow margins, including one who beat her opponent by 258 votes — after we’d put tens of thousands of video ads in front of voters in that district.

The folks at the Hometown Project did the REALLY hard work. They handled the overall logistics, collaborated with party staff to identify key districts, found potential energizers who grew up in those areas, persuaded them to get involved and got the videos shot, edited and packaged. I had the fun part: figuring out the digital strategy, launching the ads, monitoring the results and adjusting the campaigns on the fly.

Targeted Content

The sophistication I mentioned started with the videos themselves. Each was already targeted at an individual race, and we also tried to optimize them to get the most out of our advertising dollars and from the limited time each energizer could devote to the project.

  1. Energizers recorded short, camera-facing videos supporting specific Democratic nominees based on scripts drafted by The Hometown Project’s team of experienced TV/movie writers. Most energizers recorded them on their phones, giving the videos an authentic sheen that also fit with the our collective digital experience during the coronavirus pandemic.
  2. When possible, energizers recorded videos for several candidates. For instance, some had ties to an area split between more than one legislative district, while others could record content for separate state legislative and state senate races. In one state, we also ran ads helping candidates for city/county races and for Congress, and several energizers were happy to record videos for Democratic candidates at all levels.
  3. Hometown contractors and staff edited and formatted the results as video ads, with candidate and energizer names prominently featured from start to finish. Highlighting the names helped make sure that the ads could have SOME effect even if a viewer only glanced at them, a particularly important consideration for name recognition down-ballot.
  4. When possible, we distributed the videos directly to the candidates to use on their own social-media properties, though that was not always possible due to campaign-finance coordination issues.

Targeted Advertising

Even great content won’t work if nobody sees it. Fortunately, we had a large enough advertising budget on this project to reach voters in every targeted district, usually via more than one channel. Facebook gave our ads high visibility, a large audience and the chance for peer-to-peer spread. Programmatic video ads let us reach priority voters across many different platforms and in a format much harder for them to skip or ignore. Combining the two let us balance out the strengths and weaknesses of each.

For example, were able to work with lists of voters exported from the voter file in both states, but they rarely contained email addresses or mobile phone numbers. We could upload them to Facebook as Custom Audiences, but with a low match rate. In some districts, we were able to connect perhaps one quarter or one third of the voters to actual Facebook users. Also, many of our targets wouldn’t happen to be active on Facebook during the few weeks our ads were live anyway.

To expand our options, we used DSPolitical for programmatic advertising across many websites, relying on their Catalist voter data for targeting. Since we were exporting from the state parties’ voter files and comparing them to DSPolitical’s voter file, we appeared to get much better match rates on their platform than on Facebook. But some voters wouldn’t see those ads either, since they might not use the internet much outside Facebook. The two together gave us the best chance of reaching our targets.

Where we could, though, we supplemented priority-voter targeting with something a little more fuzzy. For one thing, priority-voter data models are never perfect, and if you’re relying entirely on addressable advertising, some of the voters you actually need to reach can fall outside the parameters you’ve defined. Also, data models tend to have a harder time with infrequent voters and those who haven’t been canvassed, and they’ll completely miss people who register at the last minute.

As I note in How to Use the Internet to Change the World – and Win Elections, “sometimes online outreach works best when you point it like a rifle at a particular target, but a good shotgun still has its place.” In this case, in one state we were able to work with district “shape files” provided by ActionSprout to geotarget the videos on Facebook, letting us reach a broader swath of potential voters but still stay within the right district lines. In that same state, we similarly broadened the targeting of our programmatic ads in the last days of the campaign. Combining priority-voter targeting on Facebook, priority-voter targeting on programmatic and geotargeting on both Facebook and programmatic, we had the best chance of catching people whose choices might win the day.

As you might guess, I had a blast — this project gave me the kind of challenge that makes a digital communicator want to get up in the morning. We had some hiccups along the way, of course, but in the end we served literally millions of ads at a cost-per-contact of about 1.5 cents. We’ll never know how many elections we might have swayed, but we gave those candidates everything we could. In many cases, we provided just about the only outside help they received. I’d do it again tomorrow.

Sample Videos

Hometown produced the videos in 15-second, 30-second and 60-second versions, usually depending on which scripts the energizers chose to use. For the ads, I generally picked the shortest version available, but the parties and the candidates often used the longer ones organically. Two examples are embedded in the text above, and you can check out The Hometown Project’s YouTube channel for more.

Let me know if you have questions, or if you might like a dose of the same medicine. It’s good stuff.

cpd

Leave a Comment:

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

Back Top