How big data helped secure Emmanuel Macron's astounding victory

En Marche! are expected to triumph again in this week's parliamentary elections. Much of that success could be down to three men—and a program

June 10, 2017
Macron's tech-savvy team used techniques similar to those of the Obama campaign. Photo: Wikimedia commons
Macron's tech-savvy team used techniques similar to those of the Obama campaign. Photo: Wikimedia commons

From the moment that Emmanuel Macron won the French presidency, it was clear that the parliamentary elections on the 11th and 18th of this month would be critical. Despite his status-quo-smashing victory, Macron’s powers would be severely limited without a majority in the French National Assembly. Initially, even supporters feared that he would not be able to beat political odds twice in just over a month.

Yet polls this week indicate that his new party, La République en marche! (REM), could be on the cusp of winning by a landslide, introducing a major new political era for France. As outsiders analyse how Macron is managing so comprehensively to rewrite the political rulebook, it’s worth considering three names vital to his success: Liegey Muller Pons (LMP), or, as Le Monde recently dubbed them, "Three men and a software program."

Guillaume Liegey, a 36-year-old Harvard graduate and ex-McKinsey consultant, is the ‘L’ of LMP, which models both local and national political campaigns along the lines of technology start-ups. From Berlin, where LMP is opening a new office, he declares, "Maybe I’m biased, but I think what Macron has done has never been done before... he has made politics exciting again because he has shown how somebody who has not been a longterm political insider can build a powerful movement."

LMP became involved in 2016, when Macron approached them to talk about their data-driven canvassing techniques. The discussion around big data in politics these days often heralds wary discussions of firms such as Cambridge Analytica. Yet the founders of LMP, like Macron, were admirers of Obama, who in both his 2008 and 2012 campaigns famously tapped into advanced data analytics for progressive political purposes.

Liegey himself volunteered for Obama in 2008, and it was on the campaign trail that he met MIT economics graduate Vincent Pons and fellow McKinseyite and Harvard graduate Arthur Muller. All three were galvanised by the way the vast technological and scientific infrastructure of Obama’s campaign had been deployed to reinvent the seemingly old-fashioned technique of going from door to door to talk directly to voters. In 2012, back in their own country, the three men used their experience to devise an algorithm that helped Francois Hollande to defeat Nicolas Sarkozy.

Just five years on, the surge in fake news, cyber attacks, and social media manipulation has emphasised how potential political leaders must assert themselves online not merely for self-promotion, but to face down threats from across the globe. Macron showed how clearly he recognised this when he approached Liegey, Muller and Pons before making his bid for the presidency overt. "We were all fascinated by how political parties use data to create mindsets," says Liegey.

Restrictive French privacy laws mean that political campaigns can’t target individuals in the way that you can in America. Yet LMP realised that they could get similarly exciting results by identifying key neighbourhoods for canvassing. When they founded their company, they simultaneously created a software package, Fifty Plus One, which flags up the specific characteristics of political territories that need to be targeted.

Liegey emphasises that LMP’s innovations would be less effective if volunteers weren’t actually knocking on doors. "Technology allows you to work on a huge scale, but it’s human contact that gets out the vote." What was so visionary about Macron’s approach, he continues, was the way he raised the human element to a whole new level.

It was a form of deep canvassing, Liegey explains. "It was new," he says, "because Obama’s team had only wanted to assess the degree of support for Obama. Macron wanted an in-depth diagnostic of the voters. He wanted us to have a proper conversation for around fifteen minutes and find out what people believed in their guts.

"We asked eight questions: 1. If you were to demand something of politics, what would it be? 2. Is there a cool initiative in your neighbourhood that could be expanded elsewhere? 3. What works in France? 4. What doesn’t work in France? 5. What was your best memory from last year? 6. What was your worst memory from last year? 7. What worries you about the future? 8. What gives you hope for the future?"

Before “En Marche!” was launched on 6 April 2016, Macron’s campaign technically had neither money nor supporters. "By the end of July," Liegey asserts, "they had knocked on 300,000 doors and had conducted 25,000 in-depth conversations. So we had around 1.5 million words to analyse, which is more than Tolstoy’s War and Peace [about three times more to be precise], but still human-scale."

One key aspect of big data is the speed at which increasingly sophisticated software analysis techniques are evolving. "We are on about our fourth or fifth version of our software model," Liegey explains. "We started off by looking closely at what Dan Wagner (Chief Analytics Officer for Obama’s 2012 campaign) was doing, but the data landscape in Europe and America is different."

"We decided to work with the French start-up Proxem, which has developed some of the best software for analysing conversations. Then when it came to properly understanding the subtlety of the conversation we used something that’s called a human being! But even since August 2016 there’s been a lot of progress in the artificial intelligence that can be used for this work."

Asked for tips for anyone who might have the nerve to emulate Macron here, Liegey replies, "My advice to a British politician wanting to lead a new movement is don’t be constrained by your institutions. Macron did three things very well. Firstly he had a strong cause, which was to change the whole nature of French politics by becoming President without coming from the political establishment. Second, he worked with dedicated professionals and managed to raise the money to keep that team going. Third, he had a powerful narrative, so he could mobilise enough volunteers to have conversations with millions of citizens."

Easy, then, for the average superhuman. And what of LMP’s role in the parliamentary elections? "We’re not working with all REM Candidates," he says. "A long time ago—last summer—the Socialist party contracted us, on the understanding that we could only help a few dozen candidates from other parties. We try to be non-partisan, though we only work for progressive candidates. We would not work with the Front National. And we would not have worked for Donald Trump."

This, then, is the point at which Macron must demonstrate both how much he has learnt about big data and the extent to which he can soar beyond it. Liegey is confident. "Macron is amazing," he declares. "He is good with his team, and is hungry to learn. That said, there are things I believe he can still learn. I would love, for instance, to see him make a speech as powerful as the one Obama gave in Virginia in 2008, 'Fired Up, Ready to Go!'"

What of Liegey’s own learning curve? To date, LMP have worked on around 500 small and large-scale projects across Europe. It strikes me that LMP are expanding into Germany at precisely the point when Angela Merkel has a big election coming up. "No," he laughs when asked. "This German branch is a longterm investment. Angela Merkel hasn’t been in touch. But if she wanted a conversation, I wouldn’t say no."