It’s the data, stupid: why Facebook could decide the UK general election

So, who won the UK General Election TV debates? Were you furiously retweeting every Labour supporting account that showed Jeremy Corbyn in a positive light or were you repeating Theresa May’s mantra of strong and stable leadership across all of your social channels?

You were? Great. It’ll probably make very little impact.

 Photo by Christopher Furlong/Getty Images News / Getty Images

Photo by Christopher Furlong/Getty Images News / Getty Images

I’ve written before about filter bubbles and elections — and the 2017 British General Election is a weird mixture of people both aware of filter bubbles yet still choosing to embrace them — but while many ardent Labour and Conservative supporters are Tweeting to the converted, the real difference lies in data.

And for that, we need to go back to Donald Trump.

Much has been written about Trump’s using of data, especially across social media, and his campaign’s links to big data company Cambridge Analytica, but a recent piece in the New York Review of Books lays out the Trump campaign’s incredibly detailed, granular use of targeting across Facebook.

Two quotes from the piece stand out:

“On any given day…the campaign was running 40,000 to 50,000 variants of its ads, testing how they performed in different formats, with subtitles and without, and static versus video, among other small differences. On the day of the third presidential debate in October, the team ran 175,000 variations. Coby calls this approach “A/B testing on steroids.”


“[The data] was used to create predictive models of who was likely to vote for Obama, who was not, and who was open to persuasion. It also indicated who would be disinclined to vote for Obama if contacted by the campaign. These models sorted individuals into categories — let’s say, mothers concerned about gun violence or millennials with significant college debt — and these categories were used to tailor communications to members of each group.”

These Facebook tactics are no different from any company or individual who has ever purchased adverts on the platform. Custom audiences, interest-based targeting and lookalike audiences are used by everyone from SMEs to large multinationals.

Trump is no different in that regard, although you’d like to think any digital team spending $70m on digital advertising would also be able to hone down and continually optimise towards the best performing messages.

Secondly, the modelling employed will be familiar to anyone from a marketing or data department, and is almost exactly the same reason Netflix is investing in extra Adam Sandler movies or why Amazon are so good at predicting why you need to buy an extra gold pineapple just after you’ve purchased home decorations.

And why the Conservative Party are widely reported to have invested heavily in social media advertising for this election, running a number of adverts promoting Theresa May’s leadership and attacking Labour leader Jeremy Corbyn.

Some, such as the questionable editing of video clips, sail quite close to breaching Facebook advertising T&Cs, specifically the misleading element of Facebook’s prohibited content (although, as with most things Facebook, there’s a fair amount of interpretation at play here).

But in creative terms, the message is simple: hit hard, hit often and hit consistently. But will that win an election?

The New Statesman says no. I’m not quite so sure.

A lot depends on how they cut and apply the data. Trump’s winning social strategy was not to target a broad base, which would include a large number of people who would find his views unpalatable, but by using a two-pronged approach.

Firstly, suppress voter turnout for your rival in key battlegrounds. These may not be traditional heartlands but if the Conservatives have segmented their data in the right way, they can push anti-Corbyn adverts to wavering Labour voters. The message here: don’t bother to vote, stay home, you don’t really like your usual choice anyway.

Secondly, identify voters in key battlegrounds who are likely to disagree with Corbyn and resonate with May or the Conservatives, but are either floating or unlikely to regularly turn out, hence a level of Brexit-based messaging.

This somewhat oversimplifies the strategy, but it worked for Trump and if the Conservatives have done their data analysis correctly, it could return them to Downing Street with a bigger majority than expected.

In marketing terms, this is the Protein World approach. A lot of people will vehemently object to the advert, but it helps sharpen the target audience into what they are not: for Protein World’s target audience they were most emphatically not the type of people who would protest against a slimming product and for the Conservatives their audience is not into Jeremy Corbyn at all.

Protein World, Trump and the Conservatives don’t need to be loved by everyone. They just need to inspire the right amount of people to buy their product (£2m incremental sales in Protein World’s case).

Again, we come back to targeting.

According to Reuters’ recent study, 44% of UK adults get news from social media (primarily Facebook), with 8% using social as their main news source. Work out where your heavy social usage low news consumption pro-Brexit or right-leaning audience (complete with filter bubble) resides and cross reference this with projected turnout in key seats and there’s a social advertising strategy right there. This group may even be unaware of some of Theresa May’s election wobbles in recent weeks.

Does this seem overly paranoid? Does it place too much faith in the power of advertising, data and Facebook?

Possibly. But then four years ago, Facebook was busy producing a studyshowing that newsfeeds could be manipulated to change emotions and four weeks ago it was fending off accusations that its advertising could be targeted at teenagers who “need a confidence boost”.

Ethically unsettling but undoubtedly effective.

If you can’t see or or quantify it, then it can be easy to assume it’s not happening. If I was to take my own Facebook feed as an example, I’d assume Jeremy Corbyn was growing in popularity, Theresa May was one step away from having a swanee whistle playing next to every announcement, and the Liberal Democrats adverts think that I’m the sort of optimistic remainer who would be willing to help their cause in South West London.

Knowing what I know about filters and confirmation bubbles, this should set off questions, even if the polls back up my Facebook feed. No, make that especially if the polls back up my Facebook feed.

Yes, the Conservatives may not have always got it right with Instagram and PPC adverts but if your digital budget is over £1m you can afford a little bit of trial and error.

And if you’re a wavering voter in a constituency with a potential low Labour turnout, expect to see an awful lot of Theresa May in your feed between now and June 8th.