Facebook and Cambridge Analytica: A Psychographic Fairy Story

Recently a conspiracy theory has been doing the rounds that Cambridge Analytica bought data from a UK-based researcher, and that this data was used to either elect Trump, influence the Brexit referendum, or both.

At the heart of this theory is the claim that the data downloaded from Facebook was used to build psychographic profiles on potential voters, which was then used to deliver hyper-targeted messaging.

It’s not in question that Facebook has had a bad year PR-wise, and Cambridge Analytica does not come out of this smelling of roses, but does the claim stand up to any more scrutiny than that? Could this data really have influenced the election or referendum either way?

Unlike most of the people writing about this story, I’ve worked in digital marketing for large organizations and the government, so am in a good position to say if this story stands up.

The first thing to say is that the trading of personal data online is commonplace. Load up any large publishers’ website with a special browser extension called Ghostery, and you will see certainly at least thirty and maybe as many as 100 pieces of software designed to track your behaviour and package you up for advertisers.

In the background, many are also using Machine Learning and AI to connect you to the different devices you use. Yes, this is against GDPR, but no-one in the industry really understands yet what GDPR will and won’t make illegal.

The other thing to understand is that the publishers who are now against this kind of data sharing have been quite open in their support for it in the past. The Guardian was one of the first publishers on Instant Articles, launching in April 2015. Here is its then-head of mobile and video Lee Fels:

“We are moving in an exceptionally fast-paced environment where the biggest companies, barely existed 20 years ago. So we are into a world of partnerships, as the connected future will be a world of partnerships. Hence we are experimenting with Instant Articles, and with Apple News. Partnership is not just about the on-Guardian experience; it’s an acknowledgement that there are no walls in this world, particularly as distribution and proliferation increases.”

Over the last couple of years, this relationship has soured a lot, and there is now a lot of bad blood between publishers and Facebook. Why? Because publishers have put a huge amount of infrastructure in to track audiences online and make themselves attractive to advertisers – and advertisers then go and spend their money on Facebook and Google anyway. Some estimates say that Google and Facebook account for 85c of every new dollar spent on digital.

Large publishers have a vested interest in seeing Facebook fall, especially now Facebook has come out and said that they want to refocus their News Feed away from publisher content. Every minute spent on Facebook is a minute that is not spent on a publishers’ website or mobile app.

All of this begs an even more fundamental question. Why, a as a political campaign, would you buy a two-year old data dump of data exfiltrated through Facebook’s API to purchase advertising through Facebook, when you could simply use Facebook directly?

Facebook has its own self-serve tools for targeted advertising hard-coded into its platform. This isn’t a secret. Targeting people is what Facebook does. It would be akin to walking into Home Depot and buying the parts required for a power tool, rather than just buying one off the shelf.

Now sure, you could hire a data scientist who could run some clustering algorithms and identify trends which could allow you to make some marginally more intelligent decisions about where you spend the money.

But those kinds of people don’t come cheap, a lot of other people within advertising technology are working very hard on this kind of thing without a lot to show for it (they’re the other 15c in that advertising dollar), and at most, your reach would be limited to 50m people, as opposed to the 90% of the US population you could reach using Facebook directly.

They’d also have to declare all of this stuff through campaign spending. Regulators might wonder why huge amounts of campaign money appeared to be disappearing to AWS.

More generally – there’s very little evidence that this stuff works anyway, even when Facebook has done it. Facebook has spent much of the last year fending off criticisms from the likes of WPP that it is selling digital snake oil.

When Proctor and Gamble comes to the conclusion that digital platforms can’t be trusted to sell soap powder and cuts its budget by $200m (and increases its reach in the process), why would you think it any easier to use the same tools to change people’s political beliefs.

The media have spent 18 months now looking for a magic bullet that might explain Brexit and Trump that might explain their complicity in those things (the coverage given to Hillary’s emails), people not finding Tony Blair and David Cameron convincing).

If this story makes people more aware of the data that Facebook holds on them, then that is great. Facebook could certainly stand to be more careful in how it handles user data. It’s definitely time to unfriend people who do dumb quizzes. But the idea that Cambridge Analytica swung the election for Trump is a fairy story.



Categories: Spotlights

%d bloggers like this: