Jack and Jane live seven miles from Washington, D.C., in suburban Northern Virginia. The neighborhood in which they live has been a source of “swing” votes for years and is highly prized by both major political parties. Latest forecasts have shown that in order for either party to win locally and statewide, voters like Jack and Jane need to turn out and vote for their candidates.
One of the parties hires a database company to determine what it is they can learn about Jack and Jane. The company returns with their initial report. Based on publicly available records, Jack and Jane both vote in every presidential election, and about half of Congressional elections. But in Virginia, party affiliation isn’t reported as part of the voter database. And neither Jack nor Jane vote in primaries. So the party knows they’re “reliable” voters – they just don’t know who can rely upon them. What about phone identification? No dice. Jack and Jane have a land line, but use it only for their security system.
The database company tries to glean more answers from commercial data. Among other things, they learn Jack and Jane have three kids, Jack drives a hybrid and Jane an American SUV. Breakthrough! They come back to the party and proclaim they now “know” Jack is likely a lean Democratic voter, and Jane a solid Republican. For many political candidates, that’s where the analysis would end.
There’s only one problem with this scenario. It’s wrong.
Jack’s a longtime Republican, and Jane a hardcore Democrat. It’s impossible to have a conversation with Jack without him complaining about President Obama, while Jane says, “I’ve always voted Democrat and I always will.” Using the interpretation of the data from the company above leads to wrong answers. These “wrong answers” are replicated in every neighborhood throughout Virginia, and the candidate loses by fewer than 1,000 votes.
Implausible? Not really. I know Jack and Jane – and others like them. And more than a handful of races are lost by relatively small margins.
To be fair, nothing was ”wrong” with the data. It was the interpretation of the data that led to wrong conclusions. And this incorrect methodology can wreak havoc in politics, government, and the corporate world.
Fortunately, today we have much more rich sources of data that can lead us to more accurate predictions. TV viewing habits can now be matched to online consumer profiles to let us know the household watches “Hannity” every night. Online browsing behaviors can let us know Jane gets her news from MSNBC. Donor data can tell us that Jack supports religious causes while Jane donates to animal welfare organizations.
All this data can be combined with traditional voter file records in a privacy-compliant manner to provide political marketers with a unified view of the voter. And more importantly, using big data science, campaigns can determine on which voters they should focus their efforts.
What’s more, once the dataset is established, political campaigns can reach these voters and deliver highly relevant messages to them across their TV’s, smartphones, tablets and PCs – managing reach and frequency, all while measuring the impact these messages have on brand lift (favorability) and message recall (burn-in) of a particular campaign using traditional polling or digital surveys.
It’s a brave new world in politics. As races get closer and closer identifying, persuading, and turning out swing voters will become more and more important. And addressable and accountable efforts will become central in media planning as an enhancement to traditional television and radio.
Let’s hope today’s campaign managers and consultants do, in fact, know Jack.