Big data politics: Why you can’t outrun campaigns by avoiding the TV
by Peter Bijkerk
If you think you can avoid the ceaseless barrage of political ads by merely avoiding the television at all costs and keeping your telephone silenced, think again. Politicians and their campaign managers know you too well, and they know you have to consume content somewhere. They’ll be damned if they’ll let a little thing like the internet get in the way of letting you get to know them just as well.
An ancient history lesson
We call it big data in the age of the ubiquitous web, but the truth is that politicians have been amassing attachés on potential voters for at least a decade. Some experts credit the strategy to Bill Clinton’s reelection team in 1996, which decided to focus its efforts on winning over swing voters rather than the entire electorate. George W. Bush took it even further, narrowing his focus down to swing-voter Republicans so as to maximize resources even more efficiently. All of this required a tremendous amount of data mining to figure out who these people were, where they lived and what they cared about.
Fast-forward to 2008, and the Barack Obama campaign took the process to another level thanks to its well-executed web campaign. In the process of raising about half a billion dollars online, the campaign gathered around 13 million email addresses and 5 million friends across the social media landscape. After mining this data and combining this stuff from voter-registration records, the team was able to identify the cream of the crop in terms of potential voters and deliver them a personalized campaign experience — everything from organizing rides to polling places to phone calls addressing specific points people had raised online.
You might have noticed by looking at the names associated with these efforts that running a data-driven campaign works. Managing large teams of volunteers, paying for television ads and otherwise running a major campaign is expensive business (save for those dirt-cheap and annoying robo-calls). Knowing who to target lets campaigns spend their valuable resources in the right places and avoid wasting money trying to win over individuals staunchly on the other side or unlikely to actually get out and vote.
The age of big data
As it has with business though, particularly on the web, the age of big data has changed political campaigns yet again. Heading in the 2012 election, for example, President Obama has more than 20 million Twitter followers and his Facebook page has more than 29 million “likes.” Mitt Romney lags considerably behind — with less than 2 million Twitter followers around 9 million “likes” — but still has access to a lot digital data.
Thanks to technologies such as Hadoop, campaigns are able to store more data — and more types of data (think of all that unstructured social media data) — than ever before and process it en masse. Not only does this help with targeting individuals or groups, but it could also help with longer-term strategies such as where to schedule rallies and what issues to discuss at them. As feelings toward issues change, regular analyses of ever-expanding data sets should reflect those changes and let candidates shift their strategies accordingly.
This ad created specially for you
Of course, all this additional data means more than just knowing what issues matters. It also means just knowing you, period. What you drink, the car you drive, where you shop — they’re all pieces of a digital profile that lets politicians make sure they put their messages in front of the right people. It’s a process called microtargeting, and it’s very real — especially online.
Bryan Gernet, CEO of Resonate Insights, explained to me recently how his company goes about serving billions of targeted web ads to its growing stable of political customers. In 2008, he said, neither presidential candidate spent very much money advertising online, but Resonate is presently involved in about 100 campaigns, with clients that include campaigns, political action committees and other special interest groups. At present, he said, Resonate has about 250 terabytes of behavioral and opinion data; it runs a 110-node Hadoop cluster and a 300-billion-row database.
All that data feeds predictive models based on mountains of data gleaned from surveys, web behavior, demographic information and other sources to deliver specific ads to specific web users. Resonate’s algorithms take into account 300 million “decision nodes” to in order to model all the attributes of each user, and they deliver each user the best-possible ad from a collection of a client’s pre-manfactured ads. So, a socially conservative independent will see a different ad compared to a fiscally moderate liberal or a centrist independent with a laser focus on job creation as the election’s most-important issue.
Things are only going to heat up. Gernet says he foresees candidates in future election cycles doing a lot more with custom videos targeting specific voter segments, as well as putting a much bigger emphasis on how to deliver optimal ads on tablets and smartphones.
Joe Lichtenberg, VP of edge computing for Mirror Image Internet,
sees a lot of his company’s clients implementing what he calls “dynamic creative optimization,” and wouldn’t be surprised to see this approach make its way into politics. Essentially, that means advertisers create templates that can be populated with custom content on the fly to fit each web user’s specific profile. Already, startups such as DataPop and BloomReach have built businesses around this practice for customizing web site content and ads on search engines.
Mirror Image hosts application logic for customers, including Resonate, on geographically dispersed edge servers so it can calculate incoming data and deliver content in real-time without requiring it to hit a centralized server somewhere.
Does it work?
The great thing about doing this type of targeting online is that campaigns can get pretty clear evidence of whether it works. Gernet said his company can track traditional metrics such as clickthroughs and length of time engaged in pre-roll video before skipping it and getting to the featured video. Thanks to the cookie-tracking policies on candidates’ web sites, it could be relatively easy to track whether site visits or other actions follow the consumption of personalized content. Resonate will even engage in some online polling to gauge effectiveness.
And while studies suggest that voters are turned off by the idea of being targeted — some respondents even say they’d be less likely to vote for a candidate who targeted them — it might not matter too much in the long term. For one, these personalization algorithms are complex and marketers are smart, so voters won’t necessarily know when or why they’re being targeted. Another reality is that how voters feel and how they vote doesn’t always go hand-in-hand, which is why negative ads persist despite rather widespread contempt for them.
If the data says microtargeting works, there’ll be no escaping campaign ads without a serious curtailment in the amount of data web sites, mobile apps and other platforms can collect about users. Until that happens, though, all our online activity just creates a smorgasbord of personal data upon which candidates are more than happy to feast. So, you tax-hating, gun-toting, gay-rights-supporting swing voter, get ready to know why both candidates have your back in 2016.
Feature image courtesy of Shutterstock user iQoncept.