After Elizabeth Warren was said to draw a crowd of more than 20,000 people to New York City’s Washington Square Park last week, top political number crunchers were quick to scoff. CNN’s poll analyst Harry Enten vented on Twitter, “Are we really doing this thing on crowd sizes?” HuffPost polling editor Ariel Edwards-Levy did the same: “Good morning, crowd size still is not predictive of electoral results, thank you.” FiveThirtyEight’s Nate Silver said on MSNBC that “crowd sizes are one of the last things I’d look at” when assessing a candidate’s level of support.

Warren has a lot more than crowds going for her these days. She is having her best week of 2020 primary polling yet, seizing the top spot in Monmouth University’s New Hampshire poll on Tuesday and in the Des Moines Register/CNN Iowa poll this weekend. But the crowds did come first, and despite the scoffing, they weren’t irrelevant to her rise.

Unlike data gurus, political reporters tend to think that crowd sizes really do matter, and they often write about them in breathless tones. The New York Times in particular has delved deep into what makes Warren’s big rallies so big. The Times podcast “The Daily” dissected the “Anatomy of a Warren Rally” from inside the Washington Square Park crowd, and marveled at how Warren’s followers lustily chanted “two cents” in support of a wealth tax, and how determined they were to wait in post-rally photo lines for a snapshot with the candidate. Two months earlier, the Times published a multimedia analysis of the complicated “selfie” line logistics (while quibbling that the photos are taken by a Warren staffer, and therefore, are not selfies.)

So who’s right in this Moneyball-like debate—the reporters who want to assess a candidate and her supporters in person, or the analytics gurus who scoff at the effort? Do crowd sizes matter?

Um, of course they do. No, they’re not instruments that predict the political future. But so what? Nothing is. As Warren is the latest to prove, crowds are a highly effective tool for earning legitimacy and breaking out of the pack. And, just maybe, they matter even more in the social media era, because big crowds can help campaigns build a get-out-the-vote volunteer army.

Our beloved political data nerds aren’t totally wrong, but they are overstating their case. As FiveThirtyEight’s Nathaniel Rakich wrote earlier this month, “Polls are much more accurate at forecasting elections than crowd-size estimates, which don’t tell us all that much.” But polls are not exactly clairvoyant four months before Election Day. And even if crowds aren’t predictive of who’s going to win the nomination, that doesn’t make them meaningless. It’s clear they mean something.

How could Warren be where she is today—first place in Iowa and New Hampshire, second place nationally—without her crowds? She began her campaign in the low single-digits, awkwardly apologizing for trying to prove her Native American ancestry with a DNA test, and suffering from possibly sexist speculation that she lacked “likability.” As her crowds grew into the thousands, and as her selfie lines became tests of endurance, she was presenting visual evidence that many people liked her, and then some. In the view of political scientist Rachel Bitecofer, big crowds are a way to measure a candidate’s charisma. “In fact, it may be the only way to quantify charisma,” Bitecofer says. “And charisma matters.”

Still, Bernie Sanders’ big crowds weren’t enough for him to win in 2016, nor were Howard Dean’s in 2004. The question for Warren and her crowds in 2020 is whether she can turn enough of her rallygoers into effective campaign workers who can get out the vote, in the places where she needs the most help. Not every rallygoer is positioned to do that. Warren may continue to have the biggest crowds. The more important question is: Compared to her progressive insurgent predecessors, will she have better crowds?

In a tight race, a strong get-out-the-vote operation is essential, and that requires a field operation that is intimately familiar with the local terrain. In their definitive guidebook Get Out the Vote, political science professors Donald P. Green and Alan S. Gerber wrote: “The more personal the interaction between a campaign and potential voter, the more it raises a person’s chances of voting. Door-to-door canvassing by friends and neighbors is the gold standard mobilization tactic.” Warren’s selfie lines are a digital innovation on how to recruit your supporters to engage in peer-to-peer politicking, through social media sharing of the coveted snapshots.

However, not all volunteers are of equal value. Not only does political science research indicate that friends and neighbors make better canvassers than out-of-towners, but also that canvassing is more effective when volunteers are demographically similar to the targeted voters. And even sophisticated, data-savvy campaigns can have trouble strategically deploying volunteers.

In the 2004 Iowa caucuses, crowd quality beat crowd quantity. Howard Dean had long been generating the biggest crowds that year, and then tried to win in the final weeks by parachuting in his biggest fans from out of state to canvas. Dean’s orange-hat brigade inadvertently advertised to Iowans that his surrogates weren’t their friends and neighbors.

Meanwhile, Dean’s rival John Kerry, during a period when his polls were ebbing, quietly built a team of “captains” who actually knew their fellow Iowans, which helped to propel him to a come-from-behind victory. Political scientist Thomas Schaller commented at the time that the Kerry win showed that while “there is much to be said for building support through a bottom-up, follower-based movement, there remains a premium on having a top-down, leadership-driven apparatus to harness that support.”

This cycle, Warren is trying to be the leader in both quantity and quality. Her recent big crowds—10,000-plus in Seattle, St. Paul, and New York City (the Warren campaign’s 20,000 estimate for Washington Square Park was not confirmed by city officials)—have congregated in deep-blue cities that are not part of the February voting contests. But Warren followed her flex in Washington Square Park with a healthy 2,000-person audience at the University of Iowa in Iowa City.

Where we haven’t heard reports of a strong Warren field operation is in South Carolina—the fourth February state, and the one with a Democratic electorate demographically distinct from the other three. According to entrance and exit poll data from 2016, Iowa and New Hampshire have Democratic electorates that are almost uniformly white, at least half-college educated and slightly more than two-thirds liberal. Nevada’s is not nearly as white, but is almost half-college educated and is 70 percent liberal. South Carolina is majority-black. Sixty percent of its voters lack a college degree and almost half identify as moderate or conservative.

Warren’s rise in the polls has been powered by support from stoutly liberal college-educated whites. Her two-point lead in the Des Moines Register/CNN poll is largely thanks to the 48 percent plurality she got from “very liberal” Iowans. So when Warren draws a big crowd at an Iowa university, she is getting access to potential volunteers who are, in all likelihood, demographically and politically aligned with much of the state’s Democratic electorate.

A college crowd in South Carolina is a different story. The last time Warren was in South Carolina, which was last month, she drew 925 people to the University of South Carolina Aiken, packing a gymnasium to capacity and prompting Warren to give an additional speech to the overflow crowd. But, according to the South Carolina newspaper The State, the audience was “majority white and many traveled from outside Aiken.” When Warren addressed a black church in Columbia the following day, The Washington Post reported the event was “sparsely attended.” This weekend, Warren will try again, returning to South Carolina for a town hall at the historically black Clinton College.

To win the nomination, Warren almost certainly needs to make inroads with African-Americans and non-college voters, in South Carolina and elsewhere, where bigger and bigger crowds may not her best weapon. As FiveThirtyEight’s Rakich recently highlighted, a 2018 Pew poll found that those who said they had attended a political rally in the last five years tend to be ideologically liberal and college-educated.

That may help explain why boisterous crowds didn’t help Sanders in 2016 move beyond his natural base of youthful progressives. By the later stages of the campaign, his repetitive rallies may even have created a counterproductive image of insularity. Barack Obama’s enormous crowds in the 2008 primary, of course, were an unmitigated success story. But as an African-American candidate, it was easier for him to fuse a coalition of white, liberal college-educated rallygoers with black voters. If Bernie’s “27 dollars” call-and-response wasn’t enough to grow a diverse coalition, Warren may not be able to either with her “two cents.”

In the early stages of a campaign, neither polls nor crowds are predictive of the final outcome. That doesn’t make them meaningless. Warren’s crowds don’t make her a sure winner, but neither do her new leads in the polls in Iowa and New Hampshire. And if you need to break out of a crowded field, it sure helps to have a crowd.