Donald Trump’s campaign is in peril. It’s no secret that it is low on funding and staff, but these issues are just the tip of the iceberg. The underlying problem is Trump’s refusal to run a data-driven campaign, and its effects are already being seen.

In its rhetoric, platform, and use of mass media, this campaign has shattered conventionality, largely to Trump’s advantage.

The Tyndall Report found that Trump has enjoyed more minutes of combined media coverage than any other candidate in the past seven presidential cycles, which propelled him through the primaries. As of March 15, Trump had earned over $2 Billion in free advertising, according to SMG Delta.

Much of this media coverage has played into Trump’s hand. No matter how politically incorrect or inflammatory his speech on television, his numbers only increased from the start of his campaign through his victory in the primary season. For Trump, any press is good press, no matter how controversial his statement.

But we seem to have reached a point of diminishing returns. As many journalists and commentators have noted, Trump’s campaign events are massive rallies where he appeals primarily to those who already support him. This sort of wholesale approach to driving turnout should not be dismissed, but most experts believe it should be merely one component of a larger strategy to identify, persuade, and physically help turnout voters on Election Day (or for early voting).

On a recent episode of Slate’s TrumpCast podcast, Leon Keyfakh interviewed Sasha Issenberg, journalist and author of “The Victory Lap: The Secret Science of Winning Campaigns.” A recognized expert in the field of data-driven campaigns, Issenberg predicted that Trump would fail to establish a successful “ground game.”

Issenberg concedes that Trump has a coherent theory of elections, but suggests that his theory won’t be enough to win in a competitive General Election. What Trump needs to do, Issenberg suggests, is to build the type of infrastructure that will turn out votes for him in November.

But while Trump commands the passionate support of legions of Americans who want to “Make America Great Again,” he hasn’t asked much of them. Clinton, in contrast, has an army of volunteers across the country, who actively advocate for her campaign pledges, and help potential voters to get registered and get to the polls on Election Day.

At this point in the race, she has a near monopoly on mobilization efforts. And this matters. Greatly. In the 2012 campaign between the incumbent President Obama and Mitt Romney, the Bipartisan Policy Center found that only 57 percent of registered voters actually came out to vote.

That number could be even lower this time around, a possibility that only increases the value of a well-run campaign turnout operation. Faced with what many consider a bleak choice of candidates, a Rasmussen Reports survey found that 24 percent of respondents say they will vote for neither candidate.

One extremely efficient and effective campaign in recent years was the 2012 Obama campaign, due in large part to its strong emphasis on empirical data and targeted campaigning of potential voters The Washington Post’s Dan Balz described the campaign as having “prodigious appetite for research.” To pull this off, Obama’s re-election campaign capitalized on the myriad of new internet resources and technological capabilities available to them, and it paid off.

Obama’s team tapped academics, statisticians, and political scientists to produce models and statistics to predict likely voter turnout among various demographic groups. From there, the team worked tirelessly to make sure that voters who would likely support the president were registered and had transportation to a polling place.

This is the kind of nuts-and-bolts campaigning that can make the difference between winning and losing a close race in a competitive state. The question remains whether Trump will play the campaign game, or merely continue to rely solely on the media-driven approach that worked so well in the primary.

In a May interview with the Associated Press, Trump outlined his strategy while responding to criticism for his lack of a data machine. “I’ve always felt it was overrated,” he said. “Obama got the votes much more so than his data processing machine. And I think the same is true with me.”

While Trump is right that his celebrity appeal and media coverage have worked to his advantage so far, his assessment of the Obama campaign is less than accurate. Mitt Romney’s loss to Obama in 2012 (a race Trump considers to have been a very winnable election) was at least largely a result of the Obama campaign’s sophisticated use of data.

While Trump is keen to rest on his primary laurels, his loss to a better-organized Ted Cruz in the Iowa caucuses may demonstrate the kind of problems he could face in a General Election.

Political scientist Corwin Smidt of Michigan State University recently released a poll that traced the decline in number of persuadable voters from a high of 15 percent in the 1970’s to roughly 5 percent today.

Given the shortage of these persuadable voters and the small margin for error in the election, every effort must be made to get them to the polls.

Trump not only faces harsh opposition among many in the the general public, but continues to have problems within the GOP. Politico finds that 53 percent of Republicans would prefer a candidate other than Trump. Over recent weeks, there has been increasing chatter of a move to “Dump Trump” with a coup among the delegates at the Republican National Convention.

The RNC remains a large source of support for Trump (ironically, his only hope of running a successful ground game may be the party apparatus), despite many Republicans’ dissatisfaction with their presumptive nominee. Former Ted Cruz campaign manager Jeff Roe warns about Trump’s “allergy to analytics” and points out that his rocky relationship with Republicans must be mended for him to succeed in the General Election.

There may still be time for Trump to lay the foundation for a successful grassroots campaign and establish a vote-searching machine to rival his opponent’s, if he decides to rethink his original pledge to run a campaign centered around his personality.

Trump’s choice of a running mate is likely to have a large impact on his campaign. The five people reportedly being vetted all have political experience, which Trump hopes will widen his appeal within the Republican Party and to the general electorate. With a practiced politician and veteran campaigner such as former Speaker Newt Gingrich or New Jersey Gov. Chris Christie as a vice presidential pick, the Trump campaign may begin to more strategically utilize voter data as the general election season begins.

Time is running out.