The State of Democratic Data

Winners did everything right, while losers stumbled across the finish line, right? There’s a counterproductive tendency in the world of political operatives to believe that winning campaigns did no wrong, while losing campaigns were riddled with mistakes. In reality, campaign tactics, tools, and technology operate in the margins, and many of the smartest campaigns will end in defeat, and other campaigns will end in victory despite making missteps.

To that end, the 2016 presidential campaign saw the emergence of a new conventional wisdom: Democrats, who were widely believed to possess a significant advantage over Republicans in the area of data and analytics, had suddenly fallen far behind, leading to Hillary Clinton’s defeat. Of course, when a candidate who was such a strong favorite to win falls short, it’s natural to look for excuses and cast blame. This particular narrative came without much evidence. In the end, Democrats failed to see Donald Trump’s path to victory, even though it was in plain sight all along, obscured only by the margin of error in the polls the campaign used to predict state-level outcomes.

There was also little evidence from the 2016 election that the GOP had taken a giant leap forward in terms of their data and analytics capabilities. To the contrary, the Republican National Committee (RNC)’s own models predicted a Trump loss just days before the election. The GOP’s data-ecosystem is still fragmented and marred by infighting, feuds, and mismanagement.

Of course, this isn’t to say that the Trump campaign didn’t make advances in their use of Facebook as a tool to motivate their base voters. They clearly did. But this represented a tactical advantage, rather than something indicating a leap forward in their capacities regarding data and analytics.

To be clear, simply replicating the GOP’s data ecosystem based on apocryphal stories of investments of hundreds of millions of dollars without a clear sense of where those investments could have the greatest impact, would be a mistake.

Since the 2016 election, many have relied on the false narrative of a GOP advantage in data and analytics to push for radical changes in the Democratic data ecosystem. And while innovation and investment are needed, it’s important that any solution doesn’t erode the foundation that has allowed for the growth of what is truly a robust Democratic advantage. Let’s not forget that the Democratic data advantage played a key role in delivering historic victories in the 2018 midterm cycle.

And, perhaps most importantly, any solution must recognize the vital role that the Democratic Party has played and must continue to play in the data ecosystem. Starting in 2006 when then DNC Chair Howard Dean built the first true national voter file at the party, by investing in state parties, and continuing every cycle since then, the DNC has continuously invested in growing the Democratic advantage. We must also recognize the state parties and their ownership of their voter data and the critical role they play in building and improving the voter files.

In the end, the Democratic data advantage played a key role in delivering historic victories in the 2018 cycle. A partnership, built with the collaboration of the Democratic National Committee, State Democratic Parties, and TargetSmart meant that the best political data available was leveraged to beat back the regressive Republican agenda. Was the Democratic data advantage the sole deciding factor in the 2018 blue wave? Of course not. It wasn’t even the primary factor. Just as data wasn’t to blame for the ’16 loss, it shouldn’t get the credit for our ’18 win. Our great candidates, campaigns, organizers, and organizations led the way.

Ensuring that the Democratic data advantage persists requires constant innovation. While the losses in 2016 were devastating, our team responded by working tirelessly to provide Democrats and progressives with the tools they needed to win.

Here are some stats from the ’18 cycle to consider:

851: Total number of client organizations served by TargetSmart in the 2018 cycle, including Democratic campaigns, progressive organizations, and labor unions.

32: Resistance organizations relying on our data to target their outreach and organizing efforts.

22: Organizations leveraging our national analytic install file to empower Democratic analytics and integrations. Our install file is an integral component for most models trained for campaigns and organizations during the 2018 cycle.

41: Client organizations running Facebook ads based on TargetSmart Audiences in 2018.

1,108: Polls fielded by 25 different pollsters, using TargetSmart voter data, contributing to “a super good night for polling”.

328: State voter files processed and incorporated into our national voter file in 2018.

208,397,031: Number of registered voters maintained on our national voter file.

59,179,505: Number of unregistered voting age population records maintained on our national voter file.

18: Years of comprehensive vote history (dating back to the 2000 election). Historically linked across interstate moves and name changes.

158,200,000: Number of Americans with a cell phone number matched to our national file, allowing for more accurate polling, voter models, and more effective voter contact.

108,217,900: Number of email addresses maintained on our national voter file, allowing for direct outreach and social data linkage.

259,872,429: Number of street level geocodes maintained on our national voter file.

3,478,446,167: Number of records processed with our knowledgebase-driven record linkage (“matching”) technology in 2018, allowing for advanced entity recognition across interstate moves and name changes.

42,885,018: 2018 general election early and absentee vote records compiled, distributed and analyzed at https://targetearly.targetsmart.com/. Early voters were suppressed from digital ad and traditional voter targeting projects to increase client ROI.

60: Number of individual-level classification models scored across the national voter file.

19: Number of individual-level psychographic models detailing attitudinal data on key voters scored across the national voter file.

15,665: Number of client-facing on-demand file processing workflow executions in 2018 (matching, enhancement, custom).

8,482,120: Low latency, scalable API requests in 2018 cycle.

2,282: TargetSmart Audiences shared to client organizations on the Facebook platform in 2018.

60,403,234: Facebook ad reach in the context of TargetSmart Audiences in 2018

157,912,920: Facebook ad impressions in the context of TargetSmart Audiences in 2018

30,893,572: Facebook ad impressions utilizing TargetSmart EAV segments for increased ROI in 2018

10,058,103: Facebook ad reach utilizing TargetSmart EAV segments for increased ROI in 2018

88,849,914: Number of addressable TV ad impressions (through September only) served using our voter data matched to satellite television customers, allowing for directly targeted communication.