Cardboard cutouts of Facebook chief Mark Zuckerberg. Credit: John Thys / AFP / Getty


November 18, 2019   5 mins

Thirteen academics and tech leaders signed an open letter to Google and Facebook at the beginning of November, calling for a moratorium on political advertising until after the December UK general election. I think this is a terrible idea.

Yes, such social media advertising represents the worst tendencies of politics today. But asking the world’s biggest advertising platforms to put political adverts out of the reach of impressionable voters is merely the other side of the same cynical coin that sees the electorate as lab rats, mindlessly responding to stimuli.

The role of microtargetted political advertising only came to the attention of mainstream media after the Brexit vote in 2016, and then later the same year for Trump. But these same practices have been in widespread use for over a decade: and those who used them successfully before 2016 weren’t shy about the benefits.

“President Obama’s presidential campaigns revolutionized the way technology and data could be used together to identify and speak to voters’ interests,” boasted Daniel Scarvalone of Bully Pulpit International, “reflecting the increasingly sophisticated application of digital marketing to politics.” The same tactics used by his successor, and by Vote Leave, sparked outrage.

Is this all news to you? If so, here’s a quick guide.

Step 1

Gather as much data as you can on your potential audience, by combining commercially-available data sets with public records such as the electoral roll, and with your own lists. Then distinguish the individuals worth targeting, mostly undecided voters, or lukewarm supporters who might need a nudge to get as far as the ballot box.

Step 2

Tailor your message and presentation style to your target audience, using all that data you have to profile them. Forget the megaphone, you’re trying to get an emotional response by whispering something personal in their ear.

Step 3

Test different variants of your message and then use the most effective for each segment of the audience. Digital media is ideal for this, because you can measure in real time which versions people read or watch to the end, and which ones make them click through or give you their email address.

Facebook now helpfully provides a library of UK political adverts, so you can see for yourself that Labour has six alternate versions of “Give people the final say on Brexit” – with and without flags, with and without accusations of “disaster and division”, even with and without mentioning the general election itself.

Each cost the Labour Party less than £100 and was seen by fewer than 1,000 people, because they’re posted for research, not reach. So determining which version gets more response from which types of people will direct future campaigning.

The Conservatives, meanwhile, are using “Dynamic Creative” on their “Get Brexit Done” advert — which means “advertisers upload multiple image and text options, and the best-performing combination for the audience is automatically created”. The same process is behind the multiple versions of the Liberal Democrats’ “Build A Fairer Economy” advert. Both are running on Facebook and on Instagram (which is owned by Facebook).

If you know much about marketing, you’d be thinking that A/B testing is nothing new, and that you’d be reluctant to spend thousands on an advertising campaign with no idea if it worked. But now you can empirically test whether it works on a specific target audience.

One Conservative ad pairs a video clip of Boris Johnson with a message about getting Brexit done. An alternative version with a message about investing in our NHS, schools and the police, was shown only to women aged 25-55. There were doubtless other criteria for selecting that audience, but we don’t know what they are.

Facebook has various tools for would-be advertisers, including “custom audiences” that essentially let you advertise to people who are already on your contacts list, and “lookalike audiences” who are similar in some measurable ways to people already on your lists.

By combining their own data with data bought from companies such as Experian, political parties can know who’s in your family, what technology you own, and your financial history. But they can also access details as personal as your hobbies, what charities you have donated to in the past, and what festivals you celebrate most vigorously.

Naturally these kind of measures correlate with other dimensions, such as ethnicity, religion and moral values, that might not be easy, or legal, to record directly, but can be inferred indirectly.

This is how microtargeting is done, for political messaging as well as commercial advertising. You have digital profiles in various databases, and if they match up to somebody’s criteria, you will see that advert.

But is selling a political campaign, or a candidate, really just like selling cars or clothes? True, politics has always been about persuasion, winning people over to your side enough for them to turn up and vote. But it should also be about setting your ideas up for public scrutiny and challenge.

Sending different messages to different audiences is the opposite of this. Instead of everybody seeing the same party political broadcast and reading the same manifesto promises, we’re more and more likely to get a partial message aimed at us. This goes beyond the partisanship of reading a certain newspaper, or favouring a particular broadcast channel, because instead of us choosing the media we like, we are being chosen, filtered by algorithms for how well we fit somebody else’s target subpopulation.

Politicians no longer know how to connect with us, the voters. Membership of UK political parties has declined since the 1970s, when over 5% of voters belonged to one of the three main parties, to a historic low of 0.8% in 2013 (it has risen slightly since then). The move to using focus groups and surveys, methods borrowed from marketing, precedes Facebook and the shift to life online; and so the new opportunity to understand and engage with us via social media must have been irresistible.

But the nature of this online interaction, though it appears to be personal, is in fact utterly impersonal. Matthew Rice, director of Rights Group Scotland, found that the Liberal Democrats had assigned him numerical probability scores, including a 15% chance of voting Lib Dem, a 77% chance of voting Remain in a second referendum, and forty other measures including age and first language. We are all reduced to mathematical models.

This is the real problem with microtargetted political campaigning. We are no longer seen as thinking, reasoning voters who might want to ask awkward questions, suggest our own ideas, engage on an equal footing with those who want our votes. We are lab rats in endless digital experiments, faceless data points in a numbers game.

But the answer is not to remove political adverts from social media. Calls to protect adults from misleading, or just persuasive, political adverts reveal as low a view of the electorate as the people designing and placing the adverts. Instead of engaging with voters to find out why they chose to vote in certain ways, critics point to microtargetted campaigning to explain surprising or unwelcome results.

This is just as contemptuous of the public as selling us political parties the same way you’d sell soap. If the political upheavals of the last few years reveal anything, it should be that the electorate is sick of being treated as a dumb herd.

Personalised political adverts are a lousy way to do politics. But the remedy is to revive politics as a two-way conversation in public, with voters able to question, challenge and even contribute to the politicians’ messages. And asking social media companies to protect the public from political adverts would only make that harder.


Timandra Harkness presents the BBC Radio 4 series, FutureProofing and How To Disagree. Her book, Big Data: Does Size Matter? is published by Bloomsbury Sigma.

TimandraHarknes