It’s no secret that the 2016 U.S. presidential election has been unprecedented. But what you - and most of the world - probably think of first is the back and forth rhetoric between Democrat Hilary Clinton and Republican Donald Trump, and the bizarre events along the way.
What you likely don’t know is that the role of data and analytics in this election is also unprecedented. More than in any time in history, the presidential campaigns are leveraging the vast amounts of data generated every day to segment voting audiences, solicit funds, crystallise talking points, refine angles, and better target television, radio, and newspaper advertisements - as well as those annoying things we Americans call “robocalls,” the taped telephone messages from campaigns that always seem to come as your family is sitting down for dinner. All in the name of getting through the noise to the voter who might still be undecided about which candidate for whom to cast her vote.
Four years ago, we saw the potential for data to help predict the outcome of the presidential election, so we created an analytic app that blended publicly available survey data with ZIP Code information to predict the winner (Barack Obama or Mitt Romney) of the election. And we pretty much got it right. But it was a rudimentary app that simply blended two data sources to predict the outcome - and the front end was not very pretty. Now it’s 2016 and we’ve built on the lessons learned four years ago, as well as leveraged new, available data sources and leveraged partnerships to build a much more sophisticated predictive analytics app for the election (opens in new tab).
The 2016 presidential election app attempts to predict the winning candidate within a micro-demographic area by leveraging static data from seven different data sources, including U.S. census information, county level election returns data from the 2008 and 2012 elections and voter preference data from TargetSmart. Alteryx then blends these data sources with SurveyMonkey’s syndicated polling data and enables anyone using the app to drill down into predicted outcomes in different zip codes across the U.S. as well as by different voter demographics, including race, education, age, gender, and income. And, thanks to more detailed analysis, the app can predict voter behaviour and preferences even more granularly than ZIP Code: It can drill down to the census tract level. Census tracts are typically a block or two in size, and a single U.S. ZIP Code can contain multiple census tracts.
Using Tableau visualisation (see above) and Carto mapping capabilities (see below), the analytic results are beautifully displayed for the user to review and drill down into more detail.
So, how could a political campaign use this app? One example would be determining in which communities the campaign should open local offices and conduct door-to-door canvassing rather than depend on statewide or national television advertising. Because the app drills down further than just ZIP Codes into census tract data, the campaign would be able to target very specific street blocks and neighborhoods to not only encourage voters to vote on Election Day, but also influence their votes for the campaign’s candidate, making more effective use of campaign dollars.
This kind of micro-targeting based on data analytics is also being used by political firms, such as TargetSmart and Deep Root. Using analytics, these firms can determine which issues are likely to matter to different voters in different communities, oftentimes down to individual households, and propose a variety of advertising tactics to sway voter sentiment. What’s more, they can A/B test different fliers or phone calls based on the analytic results.
What, you’re probably asking yourself, does this mean for me, an IT professional? “I’m not in politics, and I don’t live in the U.S,” you say. Well, this example of blending multiple data sources to predict future outcomes has significant implications for business, not just politics. Blending static, in-house data sources (such as your customer database) with dynamic, cloud-based data sources, like Facebook and Twitter data, customer survey information, and more, can help your business users make better, more accurate business decisions that positively impact your organisation’s bottom line. More to the point, you—in IT—can help your organisation micro-target customers down to the neighbourhood or household level and create hyper-local marketing and advertising campaigns that make more effective use of your marketing budget.
Taking this a step further, imagine if you could blend multiple data sources in a predictive analytic app and then easily share it with business users across your organisation. Giving users the ability to run powerful predictive analytics themselves is a huge win-win for your company: Business users can self-serve the analytics they need and iterate as many times as they want and you - along with your colleagues in IT - are no longer the bottleneck preventing users from making time-sensitive business decisions. Not only can this save your organisation money, it can also drive greater revenue.
Even more powerful is that your business users can share these analytic apps with partners for mutual benefit, without giving away your company’s secrets. Imagine hyper-local joint marketing programs that bring significant results to your - and your partner’s - bottom line.
So, here we are, the U.S. presidential election is finally upon us, talking about what might happen in the election and what might be possible for businesses like yours, if you adopt self-service predictive analytics like we use in the Alteryx election app. Want to know how accurate we were? How closely our predictions matched with real election results? And where were we completely off-base?
We’ll be back with part two of this series after the U.S. elections to analyse our successes and failures - and correlate those with successes and failures you might experience by using predictive analytics in your organisation.
Dan Putler, Chief Scientist, Alteryx, Inc. (opens in new tab)
Image source: Shutterstock/ESB Professional