Members of the Republican National Committee's data team agree with Hillary Clinton on this: Their data operation surpassed that of their Democratic counterparts in the 2016 election. And it served them especially well in Wisconsin, where Donald Trump became the first Republican to carry the state in a presidential election since 1984.
"Early on, we saw that Trump maybe wasn’t ahead in the state, but we saw how he could get there," said Brian Parnitzke, director of turnout and targeting for the RNC, in an interview last week. "So we knew the state was in play, and we knew that it was worth investing resources in Wisconsin, not only in the field game but the candidate's time."
Trump visited Wisconsin often, for both the primary and general election, and has continued to return for rallies, official visits and fundraisers since he was elected.
The president's victory margin in Wisconsin was thin, but it was enough. It also came as a surprise to those who had followed public polls that consistently predicted Clinton would win the state.
The former Secretary of State did not visit Wisconsin after losing the state's primary to Vermont Sen. Bernie Sanders. Trump also lost the state's primary, to Texas Sen. Ted Cruz.
It wasn't just Trump that surprised political observers. Republican Sen. Ron Johnson defeated Democratic former Sen. Russ Feingold for a second time, by 3 percentage points, and Republicans gained historic majorities in the state Legislature.
RNC operatives credit those wins in part to a robust data program that was first prioritized under former RNC chairman (and former Republican Party of Wisconsin chairman) Reince Priebus, now Trump's chief of staff. After suffering blistering losses in 2012, the party has since invested $175 million in data, digital and field operations.
"In Wisconsin ... we were able to find those areas that were most open to Trump’s message, and it worked," said Mark Jefferson, director of majority retention for the RNC and former executive director of the Wisconsin GOP. "And Ron Johnson used it effectively as well. That campaign was tuned into this as well as any in the country."
The data program was especially helpful in October, when the party noticed that Trump had lost ground in Wisconsin compared to the previous month — and specifically targeted voters who had drifted away in an effort to bring them back, Parnitzke said.
That wouldn't have been possible without a "voter scores" data operation that has been built, expanded and refined over the course of each election since it was launched after Democratic President Barack Obama's re-election in 2012.
Watching the way Democrats made use of data and digital operations in Obama's elections gave Republicans a sense of urgency, Jefferson said. Although Republicans have seen far more success in 2014 and 2016, the urgency is still there, he said.
"The electorate is changing. People's habits are changing," Jefferson said. "You have to keep up or you're going to be left behind."
Clinton, in an interview with Recode's Kara Swisher last month, was harshly critical of the DNC's data operations. The DNC's data was "mediocre to poor, nonexistent, wrong," she said.
Democratic data scientists and analysts were quick to turn the blame back on Clinton and her campaign. The problem wasn't the data, they said, but how it was used.
Newly-elected DNC chairman Tom Perez told CNN's Erin Burnett earlier this month that the DNC has to "up our game" and is investing in technology, training and organizing.
In the meantime, RNC operatives are encouraged by the results they've seen in recent special elections in Montana and Georgia, where they say their models were not only spot-on, but enabled them to target undecided voters to push their candidates over the edge.
"These models that we build, they’re predictive, but they’re also prescriptive," said Conor Maguire, RNC director of external support.
Maguire, Parnitzke and Jefferson were in Wisconsin this week training state party operatives on the program. The data it contains can be scaled up to the presidential level but down as small as a city council race, they said. Models will be built for Gov. Scott Walker's likely re-election campaign and for the Republican who challenges Democratic Sen. Tammy Baldwin — but they can also be used for state judicial and legislative races.
"They’re useful to the party between now and election day in 2018, but there’s a lot of things going on between now and then," Jefferson said.
The party has already used data indicating 75 percent of Wisconsinites want Democrats to "find a way" to work with Trump in an ad campaign urging Baldwin to hold a vote on Supreme Court Justice Neil Gorsuch's nomination, Parnitzke said.
Similar work will be done in states throughout the country, but Jefferson noted that Wisconsin has the unique benefit of a strong connection between the national party and Gov. Scott Walker's campaign — an operation whose data program has earned national acclaim and has benefited Republican candidates up and down the ticket in the Badger State.
"If you have an 'R' next to your name, you have access to our data and all of our resources," Parnitzke said. "That’s our only agenda. We’ll train people up on how to do this. We don’t have some proprietary secret sauce that we are unwilling to share with people. We want to evangelize and spread this as far and wide as we can within the Republican Party."