What do journalists get wrong when reporting on Facebook advertising?

Reporters are right to hold Facebook to account so political influence isn't brushed under the carpet.

Unfortunately, reporters are also missing some facts about how the platform works, and jumping to conclusions about what's really happening. This is totally understandable - it can seem like a complex area, and you have to know the details in order to ask the right questions.

I want reporters to get this right, so we can all make better judgements about how these tools might be used against us, for political gain or otherwise.

Let's pick apart one recent story to see where the assumptions are made, and try to fix them.

At the weekend, the Observer published a piece titled "Revealed: Tory ‘dark’ ads targeted voters’ Facebook feeds in Welsh marginal seat".

The ultimate point it makes is that political advertising is completely unregulated. That's fair, and I'd agree there should be rules of engagement, albeit based on the facts on how these ad platforms work.

Let's set the scene then pick apart some assumptions made in this post:

Charlotte Gerada, who works for a charity, told the Observer that she and a friend had raised £13,000 in a few days with the idea of increasing the turnout among 18- to 25-year-olds in key marginals.
“We looked at the seats where we thought we could make the biggest impact and created materials aimed at encouraging young people to vote. We calculated that if only half of those who clicked on the adverts decided to go out and vote Labour, we could still make a real difference.
“We used Facebook’s ‘lookalike’ audience tool to find young people, and initially our adverts in Delyn were both our cheapest and our best-performing. But the day before the voter registration deadline, the price of the ads shot up. We had been paying around £1.08 a click before and it spiked to more than double that. At one point it was up to £3.40 a click. The way that Facebook ads work is that you ‘bid’ for the slots... so if the price shoots up, it means someone else is bidding against you for the same slot.”

The facts so far:

  • They were targeting 18-25 year olds in specific locations
  • They used the lookalike tool, which finds the top 1-5% of people in a country (not specific locations). Lookalikes are inherently based on some kind of source audience, which isn't detailed in this post. (In our case, we usually feed the lookalike tool with previous customers to find the top 1% of people just like them, and automatically update it with every single sale.)
  • They were paying £1.08 per click. This is really high for Facebook. Our good ads get under 10p per click and a 'bad' one might be getting 40p. Some ads aren't meant to get clicks, which can drive the price up - for example, you might be aiming for video views, which are great for awareness but bad for clicks. In this case, you might be paying pounds per click, and that's fine if your goal isn't to get traffic.

The glaring assumption:

The way that Facebook ads work is that you ‘bid’ for the slots... so if the price shoots up, it means someone else is bidding against you for the same slot.”

Wrong. Almost no-one sets a 'bid'.

Facebook technically bills for impressions, and it's absolutely routine to use the auto-bid option. The old rumours that Facebook takes more money than it deserves were stamped out long ago when we all realised it works in our favour.

Further, there's a lot of ad slots available. People will be scrolling through a feed and might see several ads. This is why Facebook bought things like WhatsApp, which gives them far more inventory and therefore a tonne more slots for ads. And on top of that, finding new ways to insert ads, like in the middle of instant articles, with new video formats, and so on.

You won't seen one advertiser taking up multiple slots. If the Labour ad was pushed out, it could be any one of a hundred advertisers taking those positions.

So on their bidding, one of the following must have happened:

  1. They set a bid (which you can set for clicks or otherwise, but in reality Facebook applies it to ad impressions), and the moment any other advertiser uses auto-budding, they come out on top. If this is the case, the issue is the original advertiser's naivety.
  2. They had auto-bid on, but their ad was genuinely less engaging to the audience. If so, this says more about where the target audience's interests and allegiance lies, or that the advertiser simply did a bad job of making an engaging ad. Often, it's as simple as the ad not asking the viewer to take action. "Click here to find out more" can transform a campaign just by asking the obvious question.

Now from the Facebook expert:

Quentin Johns, a Facebook marketer who was advising the campaign, told them to set up dummy accounts with the same profile and geographic location of the voters they were targeting to try to figure out who was competing against them. In Delyn, they discovered ads from the Conservative party were being displayed in place of theirs.
Johns said: “There’s only limited ad space on Facebook, so if you’re targeting a particular demographic in a particular area and have a lot of money, you can simply drown out other advertisers. And that’s what was happening. The ads encouraging people to vote were wiped out by these adverts.”

I don't mean to question Johns' experience or ability, but have to point out a couple of issues here.

The dummy accounts with the same profile and geographic location won't have the same history of engagement as the other 'normal' people any of these ads were targeting.

If you run ads to a lookalike audience, Facebook is finding people with the interests and behaviours of whatever source audience you fed into the lookalike in the first place.

Johns' recommendation to set up dummy accounts would only be effective if, say, you ran ads with absolutely basic targeting: people of a specific age who live in a specific location. They've already established that they used lookalike audiences, so Johns' test is dud.

His quote is technically correct; "if you’re targeting a particular demographic in a particular area and have a lot of money, you can simply drown out other advertisers."

But that's not what they did. They used a lookalike audience, not a particular demographic in a particular area, according to the article. And anyone who only uses Facebook to do such rudimentary targeting is never going to succeed, unless they're giving away free beer.

We need to know more about the source audience they used to feed the lookalike audience, or if they used any other targeting options, before making these claims.

Subscribe to Tom Davenport

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.