Total Pageviews

Wednesday, June 19, 2013

From Campaign War Room to Big-Data Broom

At the Obama 2012 campaign headquarters in Chicago, inside a room nicknamed “the Cave,” a team of data analytics experts helped President Obama build the most technologically advanced campaign in American history. And in their own centers of power, even before those votes were cast, Republican leaders and strategists vowed to hone their high-tech skills for future Election Days.

In ways that are often invisible to voters, Big Data â€" the suite of powerful technologies and digital measurements that has upended so much in the world of commerce â€" is reshaping American politics.

This is the new electioneering. Campaigns analyze data like voter files and buying habits to pinpoint potential supporters, donors and volunteers and, crucially, to marshal votes. Political advertising, like all advertising, is increasingly tailored to a particular person’s interests through the use of digital information and computer algorithms.

Part of this new world is unfolding in Blue Bell, Pa., near Phladelphia, where a small company called BehaviorMatrix is trying to sharpen the Republican Party’s technical prowess.

The company’s co-founder and chief technology officer, Charles Davis, has helped develop a set of technologies and algorithms that quantify and measure voter emotion and opinion online. The company is tracking what people are saying on Facebook, Twitter, blogs and elsewhere to determine how people think and feel about issues that might matter at the polls.

Whether Republicans can close or even reverse the Democrats’ current edge in data analytics in time for the 2014 midterm elections â€" and, of course, the 2016 presidential campaigns â€" is uncertain. But BehaviorMatrix has already teamed up with a digital shop, CrowdVerb, on behalf of South Carolina Republicans for work on a special Congressional election this year. They are also working on the re-election campaign of Senator Mitch McConnell, Republican of Kentucky and the Senate minority leader.

Data-driven! campaign techniques like microtargeting â€" using data mining to deliver a tailored message to a specific subgroup of the electorate â€" cannot change the fundamentals of an election or make an unlikable candidate suddenly likable. But the value of every vote became clear to political operatives after the hair-thin margin in the 2000 presidential election. During the last several election cycles, microtargeting has become a critical tool of politics. Campaigns use voting history as well as demographic and consumer data to determine who is likely to vote, and for whom, and direct customized advertising messages and fund-raising appeals to those specific voters.

“We can pretty accurately predict who is going to vote based on what they’ve done before,” said Rayid Ghani, the chief scientist of the Obama 2012 campaign’s data analytics team.

Mr. Ghani says adding social-media data like someone’s Twitter feed probably would not add much insight. But BehaviorMatrix and CrowdVerb are betting tat campaigns will want all the real-time data they can get.

Until now, most microtargeting was based on backward-looking data â€" the car someone bought or the magazines they subscribed to, says Cyrus Krohn, a co-founder of CrowdVerb. But as more people communicate online, particularly via social media, the Web has become an alluring source of up-to-the-moment insights.

“The social angle is clearly where this is going,” said Patrick Ruffini, a Republican strategist and president of Engage, another digital agency. “There’s only so much you can do with ‘So-and-so drives a Volvo.’ ”

Mr. Davis said he had an idea of where things might be heading in the late 1990s, while he was chief technology officer at Millward-Brown Interactive. “It was clear to me,” he said, “even at that early stage, that the Internet was an evolving socio-behavioral phenomenon and that new tools and methods needed to be developed to effectively measure trends and attitudes.”

Over the ne! xt decade! , Mr. Davis established the intellectual foundation that would underlie BehaviorMatrix, and in 2008, while seeking start-up funds, he met William M. Thompson, the former chairman and chief executive of Innovative Tech Systems, a software developer. Mr. Thompson invested in BehaviorMatrix and came out of retirement to run it.

Other data scientists and mathematicians at BehaviorMatrix built on Mr. Davis’s work, and the algorithms they use today try to analyze online speech and interactions, quantifying how people feel about a candidate or issue and building a model to predict how individuals may vote or whether they will buy a product. By analyzing the interplay of networks online, BehaviorMatrix also seeks to identify which individuals are influencing the opinion of others, Mr. Davis said.

BehaviorMatrix has done work for corporations and governments, but in its partnership with CrowdVerb it has dived into electoral politics. And Mr. Davis, 41, who grew up in a family of Democrats in the libera San Francisco Bay Area, has found himself in the odd position of working to improve data analytics for the Republican Party.

“Does it make for some interesting conversations when I’m back home at Christmastime?” said Mr. Davis, a registered independent. “Sure it does.”

While BehaviorMatrix has measured and modeled online opinion and emotion for clients before, it has just begun trying to link online profiles with flesh-and-blood voters.

The company’s algorithms also use publicly available information, like a person’s name and address, to tie a name in a voter file to a comment on a blog or Twitter post.

Tying the two identities together “allows you to use data you’ve collected about someone online to model offline behavior, and vice versa,” said Sasha Issenberg, a reporter at Slate and the author of “The Victory Lab,” a book about the science of campaigning.

This is what CrowdVerb and BehaviorMatrix did during the special election in May for the ! First Con! gressional District in South Carolina, which Mark Sanford, whose tenure as state governor was nearly ended by scandal, won handily.

The South Carolina Republican Party hired the companies to check voter files. BehaviorMatrix ran the files against a database of online profiles and pulled a sample of more than 4,000 matches, weighted to represent the district’s electorate, Mr. Davis said.

Among those matches, discrepancies were found between what BehaviorMatrix predicted a voter would do based on what that voter was saying online and what the traditional identification work predicted about that voter.

It is not yet clear whether the data CrowdVerb and BehaviorMatrix generated was more predictive than the traditional voter file. Once the South Carolina State Election Commission releases the voter rolls, CrowdVerb will do follow-up interviews to check if the digital profiling was better than the traditional voter profiling, and if so by how much, Mr. Krohn said.

But some wonder if addig online data to voter files will prove hugely beneficial.

“I don’t think the next big step will be about a new source of data,” said Joseph Rospars, the Obama campaigns’s digital strategist and co-founder of Blue State Digital. “It will be about better understanding the data we have.”

Still, Republicans are eager to expand their technological tool kit, and BehaviorMatrix and CrowdVerb have already moved on to Mr. McConnell’s campaign in Kentucky. Even if Mr. McConnell does not face a top-tier challenger, either in the Republican primary or the general election, Mr. Davis said he hoped the technology he helped develop will prove beneficial beyond who wins and loses, or even righting the Republican’s technological ship.

“This isn’t just about figuring out who’s going to vote and how they’re going to vote,” Mr. Davis said. “One of the most important things that the candidate does when they use our systems is they actually understand why people are voting fo! r them or! why they’re not, and that has the effect of hopefully being able to change policy in a more meaningful and democratic way.”