We've built a data-machine
Democrats have spent years telling a story of technological sophistication: enormous voter databases, predictive models, and campaigns equipped to identify exactly whom they need to reach. In 2019, the Democratic National Committee announced a new data warehouse, calling it a “game changer.” By August 2020, its technology team reported managing more than 70 billion model scores covering over 200 million registered voters. These were tools intended to guide actual campaign decisions—whom to contact, recruit, and turn out.
The promises went beyond size. In July 2020, the DNC said a new texting model could make successful contact up to four times more likely. Its broader voter-data operation, the party explained, would help campaigns reach “the right voters” with “the right message at the right time and the right place.” That is a claim of precision. When it comes to winning elections, though, who actually gets contacted deserves more attention.
Razor-thin elections
In 2020, Joe Biden won Arizona by 10,457 votes and Georgia by 11,779—margins of approximately 0.31% and 0.24%, respectively. If percentages obscure the reality, think of it this way: Michigan Stadium holds more than 107,000 fans. Georgia’s margin, scaled to that crowd, would amount to a winning edge of roughly 250 people in the entire stadium.
Congressional races can be even closer. In 2022, Republican Lauren Boebert won by just 546 votes, a margin of approximately 0.17%. In 2024, Democrat Adam Gray won by 187 votes, approximately 0.09%. In elections this close, tiny changes in persuasion or turnout make data-enabled microtargeting consequential.
So who is getting targeted?
These razor-thin margins would seem to make reaching potential voters outside the committed partisan core especially important. That includes independents. But identifying people genuinely open to persuasion takes more than a party label - precisely the kind of challenge a sophisticated data warehouse should help address.
Pew’s research on the 2020 election found a clear relationship between political involvement and reported campaign outreach. Highly engaged voters - those who donated, volunteered, or attended campaign events - reported more forms of contact. About half said campaigns reached them through at least four channels during the final month. By contrast, voters who had done none of the political activities surveyed reported the fewest contact types among voters. Among nonvoters, 30% reported receiving none of the six forms of outreach surveyed; another 20% reported only one. These figures measure channels, not the number of messages, but they reveal an uneven distribution of attention. The findings cover campaigns generally, not Democrats alone.
The pattern predates 2020. In a Pew survey conducted in August and September 2016, 49% of registered voters who had voted in 2012 reported campaign contact, compared with just 34% of those who did not vote. Before the final push even began, people with a record of participation were more likely to have heard from a campaign.
There is a defensible reason for prioritizing people already involved. Past voting provides campaigns with a signal about future participation. Campaigns are also fast-paced projects with limited resources. People who usually vote can appear to be a better investment than those who rarely do.
But that logic risks reinforcing existing disengagement. When past participation determines future attention, people with thin voting histories can remain outside the conversation. A prediction that someone is unlikely to vote does not answer a different question: could effective outreach make that person more likely to participate?
That is why the standard for evaluating campaign technology must extend beyond database size, model counts, and total messages sent. Those numbers describe capacity and activity. They do not establish whether campaigns are reaching beyond the people already inclined to participate.
Democrats’ claims of extraordinary targeting power invite a straightforward question: how much of that power is being used to reach people beyond committed believers and habitual voters?
Having everyone in the database is not the same as bringing everyone into the conversation.