WASHINGTON -- The Democratic National Committee is using large-scale data on voters -- including a collection of millions of cellphone numbers -- and applying data science techniques to refine how the party reaches out and digitally knocks on voters' doors this fall.
The DNC announced last week that it had updated its national digital voter file of more than 300 million Americans with an infusion of millions of cellphone numbers it bought from a third-party vendor in January. The DNC then applied its proprietary data science model, named Sonar, which it has been building in the past few years, to the database and refined it so when campaigns call a voter they can be sure to connect with the person they intended.
Although the national database had been in use before, a new element is Sonar's ability to predict which phone numbers are accurate and how voters prefer to be contacted, said Nellwyn Thomas, the DNC chief technology officer. "This will absolutely be used by the Biden campaign in the presidential race but also in the House and Senate races," she said. Former Vice President Joe Biden is the expected Democratic nominee.
Political campaigns are "effectively giant communication engines and they're communicating at many different levels" with voters, Thomas said. Campaigns use a combination of approaches to drive their messages, such as large gatherings, news events, announcing policy proposals and direct one-on-one contacts. But door-to-door canvassing "doesn't scale as well, especially in an era of COVID, so using phone calls and text messages becomes increasingly important," she said.
The push to sharpen digital tools comes as Democrats and Republicans vie to reach voters during a pandemic that has halted traditional ways of campaigning, including in person fundraising and knocking on doors. Donald Trump's successful 2016 campaign leaned heavily on a digital strategy of reaching small, targeted groups of voters through social media platforms.
Thomas, who worked for the Hillary Clinton campaign in 2016 and later for Facebook, became the chief technology officer at the DNC in May 2019. Since losing the 2016 race, the DNC has beefed up its data operations and launched a new platform using Google's Big Query, a cloud-based data warehouse, as the main repository. The DNC voter database is named Phoenix, after the Greek mythology of a long-lived bird that rises from the ashes of its predecessor, Thomas said.
The DNC has invested tens of millions of dollars in the digital effort and has hired about 60 data scientists and technologists from several tech companies to run its operations, with a focus on privacy and data security, Thomas said.
The goal of the data-science driven voter file is to give volunteers the best and most accurate way of reaching voters, Thomas said. In the 2020 cycle the DNC has purchased as many as 60 million cellphone records from vendors who specialize in such datasets and sell them to companies and others seeking to digitally target consumers.
A voter who first registered in Michigan 25 years ago and listed a landline number on their record, for example, may now use a cellphone exclusively, and campaigns need to know that, Thomas said.
Also, if a field organizer in New Hampshire has 10 volunteers to call voters on a Sunday, for example, and they can reach 500 voters in two hours, it's key that their time is not wasted on disconnected numbers, calling people who have never voted, or dialing registered Republicans, Thomas said.