What Google’s targeting ban has meant in practice

In late 2019, Google banned some types of targeting for political ads. Specifically, they stopped allowing ads using ‘interest-based’ and ‘custom audience’ targeting (where the campaign collects or profiles a list of people, matches them to users on that platform and who are then shown ads). Instead, Google now limits political ads to targeting by age, geography (postcode, not radius), gender and context (such as a search result, the largest part of their business).  

The result is that Google has forced many ads (particularly YouTube ads) to be more ‘broadcast’ than ‘microtargeted’. But in doing so they’ve made these ads useless to campaigners looking to get potential supporters to take action.

To go into more detail, the upside of Google’s policy is that there’s little risk of campaigns abusing supporter data. Without ‘custom audiences’, there’s no way for a campaign to use lists of supporters/profiled voters to run ads and, as such, there is no benefit to collecting this type of data on voters’ behaviours and preferences in the first place. This means there’s one less way for campaigns to run voter suppression/demotivation campaigns targeted at specific groups (though campaigns can still target zipcodes with a higher proportion of minority voters or women or younger voters).

The downside is that campaigns who want to reach people who already support them can’t advertise to reach specific groups who might be genuinely interested in their efforts. This forces a ‘broadcast’ approach, making the campaign cast the net wider to reach ‘their’ people. Because a campaign is more likely to use a more general message when doing this, it likely reduces their effectiveness when it comes to getting supporters to participate.

You can see how this plays out in practice by looking at Joe Biden’s ads in the Google and Facebook political ad libraries.

On YouTube, everything he’s running is similar to a TV ad, designed to have a fairly broad appeal and not to be jarring if you aren’t part of the actual target audience. Through them, we can see how he’s trying to reach groups of voters who make up his ultimate electoral coalition by targeting them where they live (essentially using an urban vs. rural split based on concentrations of ‘his’ voters).

How a Biden YouTube ad targets Pennyslvania

The context matters too. These ads are seen by people before they get to the videos they actually want to watch, so the campaign doesn’t even to try to generate a click from them. Overall, this means the ads will get little or no direct response, and overall participation in the campaign resulting from YouTube ads will be lower.

On Facebook and Instagram by contrast, you see more ‘direct response’ ads, resulting from the continued ability to use ‘interest’, ‘custom audience’ and ‘lookalike audience’ targeting. These ads try to get clicks for specific actions – sign ups, donations and volunteering – the campaign wants people to take. And though Facebook doesn’t routinely publish targeting data in their ad library (they should), the content of the ads shows us how the campaign is working to reach and motivate narrower, more specific groups of voters. 

A Biden ad on Facebook, soliciting a direct response

So which is the right way to go? As with many of the issues Who Targets Me covers around political ads, there are multiple trade-offs.

Google’s removal of ‘interest-based’ targeting and ‘custom audiences’ costs them some advertising revenue (though it’s not as big a part of their business as it is for Facebook), reduces the risk of them getting caught in a future political data scandal such as Cambridge Analytica and gives them some positive ‘doing something’ PR. But this comes at the cost of many legitimate, often smaller, campaigners being able to reach the people they would like to reach.

Facebook, on the other hand, retains the income from accepting all types of targeted advertising (particularly lookalike and custom audiences, which are a very large part of their revenues), though keeps open the risk of another future crisis relating to the misuse of political data.

It’s not obvious what the right answer to this trade-off is, but it seems to us that neither platform has got it right. 

Google favours big, wealthy campaigns with the resources to put into general awareness raising and persuasion, whereas Facebook covers the entire spectrum of campaigning, without sufficient transparency and safeguards against the abuse of political data.

We believe a better answer lies in ensuring radical transparency for political ad targeting, strong measures against voter suppression (which the companies are getting better at, though primarily in the US, where the public relations risks are highest) and real confidence that any political data uploaded to a platform was legitimately acquired and properly treated (we propose regular random data audits of large political advertisers to ensure compliance).

Regulation would achieve this, but the fact that the two largest platforms now have different policies on targeting could also have been reconciled by one simple act –  talking to one another and agreeing a shared solution.