To subscribe to the monthly C&E Europe email newsletter and event announcements click here.

Several weeks before the U.S. presidential election, the Clinton campaign invited Democratic big city mayors from around the country to its Brooklyn headquarters. Top campaign management gave a lengthy presentation explaining the impending victory with elaborate charts, graphs, and detailed demographic segmentation in states and urban areas.

Thoroughly confused, one mayor from the Midwestern U.S. raised his hand and asked, “Let’s take a break from the data. Instead, could you explain your message to working class voters in sixty seconds?”

The campaign advisors struggled for several minutes. “Thanks,” said the mayor, “based on what I’ve just heard, I’m here to tell you that you’re in serious danger of losing my state.” Though two other mayors immediately endorsed their colleague, the campaign never spoke with them again.

For the last three presidential elections, American Democratic political operatives have been the world’s most advanced campaigners. It’s why there is increased interest around the world in U.S. campaign tactics, particularly those employed by the Obama campaign in 2008 and 2012. In 2016, the Clinton staff constructed a powerful mathematical model to predict voting patterns and necessary victory margins in swing states throughout the country. It was supposed to be the most sophisticated, technical campaign in the history of American politics.

Looking at Donald Trump’s campaign, we Democrats laughed at reports of his miniscule “ground game,” with only a small percentage of field offices running door-to-door-canvassing. We were almost embarrassed that he continued to hold rallies in states he was already sure to win. And we scorned Trump’s racism and misogyny, sure Americans would reject a man so unfit for the Oval Office.

Yet somehow, Democrats came up short. So what happened? And what can European campaigners learn from it?

Yes, Secretary Clinton won the popular vote by nearly three million ballots and two percentage points, but those are hollow victories. Thanks to the peculiarities of the United States Electoral College, she won her votes in the wrong places, sweeping states along the coasts by significant margins and losing crucial swing states in middle America by perilously close margins.

It’s these razor-thin losses in places like Michigan that nag progressives: the Clinton campaign model predicted she’d win the state by five points. The devastating loss has lead progressive campaigners across the world to ask a fundamental question: Does professional, data-driven campaigning really matter?

It’s important to put data-driven campaigning in context. At its most basic, American operatives run “Get Out the Vote” mobilization campaigns, combining publicly-available information from the voter registry and purchased consumer data to identify the party’s base supporters who, for one-reason or another, don’t vote very often. The campaign sends volunteers to these voters’ homes to encourage them to go to the polls.

Even with sophisticated targeting models of infrequent voter segments, academic studies of Get Out The Vote plans estimate even the most effective may increase turnout by seven percentage points. While seven points can make a difference between winning and losing, it cannot make enough of a difference that a party should rely on data alone.

Though Donald Trump committed many unforced errors as a candidate, his message unquestionably resonated with working class voters in the Midwest: You’re struggling to make ends meet because our government is broken and has forgotten about you. I will fix it. They overlooked his racism and misogyny because his distain for political correctness actually reinforced his outsider status.

Just as the Brooklyn meeting exposed, Hillary Clinton did not have a competing economic message for working class voters across the Midwest. She failed to empathize and identify with them. Not enough heard her say, “I know that we’ve come a long way in the last eight years, but I know too many Americans are still struggling.”

When she did talk about pocketbook issues, she tailored her message to specific interest groups her campaign sought to mobilize. She spoke about affordable university education to motivate young people to the polls; she spoke about paid leave from work to motivate women. Both messages are extremely important, but both were designed to reinforce the data-driven mobilization plan as dictated by the mathematic model.

In other words, message became subservient to data. It was a critical error. Without a compelling, broad-based argument, Trump stole a crucial part of Democrats’ base. Democrats typically win 60 percent of union voters. In 2016, we won 50 percent.

As digital campaigning spreads throughout the world, campaigners in Europe should embrace it, but be careful to avoid the trap the Clinton team fell into: Do not forget to persuade voters first with a broad economic case as the party’s core message. Data-driven does make an important difference in competitive races, but only if it supports the campaign’s message, never the other way around.

Jim Arkedis is a progressive political consultant who specializes in bringing American campaigning strategies to international clients. Follow him on Twitter @JimArkedis