Skip to main content

How Data Science is Used to Target the Right Audiences

Before mass communications data was readily available, communicators and advertisers had to take a shotgun approach – spread your message far enough, wide enough, and loud enough, and eventually it’ll reach some of the people that you intended.

But today, the number of people and the number of messages has increased exponentially, meaning that a more targeted, strategic approach is needed if your message is to get through. Fortunately, data science is making that possible.

Predicting Voter Outcomes

In 2010, just two years after Barak Obama’s election as president, Democrats lost control of the House of Representatives. It was a demoralizing defeat, but not an unexpected one: Dan Wagner, the DNC’s targeting director, had seen it coming.

When Wagner was hired in January 2009, he began working with the DNC’s technology department to develop Survey Manager, specialized software that turns voter survey data into tables and predictive results.

As the 2010 midterms approached, Wagner discovered that individuals surveyed via phone were much less likely to vote than previous, conventional models would have predicted. Those models had relied upon analyzing small samples that were then treated as representative of the whole. Wagner instead thought of voters as individuals. Using his new techniques and Survey Manager, he predicted the outcome of each congressional race—and he was off by only 2.5 percent.

Today, analysts like Wagner start their work with a voter database that includes registration information, along with details of whether and when a person voted. From there, the analysts append supplemental information from commercial and other data sources that could include demographics, occupation, political contribution history, magazine subscriptions, political volunteer history, and much more. And finally, individual campaigns now begin surveying voters and adding very specific information to the database. Questions could range from who you intend to vote for to your views on a specific issue.

And while Cambridge Analytica, which sold its services to the Trump campaign, is the most well-known for this type of work, they’re hardly the only ones doing it. The Democratic National Committee has used Catalist, a database with info on over 240 million Americans, for the same purpose.

More Scientific Hiring

For every corporate job position advertised, an average of 250 résumés are submitted. And if the position is listed on an online job portal, employers could receive hundreds of thousands of résumés to wade through. Sometimes the person sifting through these isn’t even the hiring manager—it’s a recruiter or other third party just trying to make sense of it all. And if they choose poorly, a bad hire can cost a company $50,000 or more in a year.

With all the information out there, and the financial importance of making a good match, how can a company make the right decision about who to hire?

Enter big data.

In the future, analysts will not only be able to use data to narrow down a list of applicants based on their résumé data, they’ll also be able to review info from social media and online assessments—then compare that to a profile of the most successful company employees to date to find the right match.

For example, look at Xerox Corp, who tended to hire call center staff with previous experience, expecting they’d be more likely to succeed in the role. For each new hire, Xerox spent about $5,000 on training, yet many new employees would leave before the company could recoup those costs.

They turned to data to give them some answers. The numbers showed that new employees with no prior experience were just as effective and stayed in the role just as long. They found that the people who stayed with the company the longest were those who were active social media users and those who were creative types. When the company began using these hiring criteria, they cut their call center attrition rate by 20%.

Customized News Feeds

At the top of your Facebook feed is a post—chosen among the thousands that your friends, companies you follow, groups you belong to, and pages you’ve liked—that the algorithm suggests is most likely to make you smile, like, laugh, share, cry, or comment.

For every post, the algorithm predicts how likely you are to interact with it based on your past behavior. What do you read, like, click on, or ignore? Using machine learning, the goal is to feed us what’s relevant and what we want to see. But what if people click like on headlines they like, even if they’ve never read the articles? Or what if they click on an article but end up not liking it? The result optimizes our feed for virality and quick interactions, rather than quality information and true connection with our friends.

And this algorithm has power, shaping what more than 1 billion daily users see and read and know each and every day. It can be adjusted to make us well-informed or show us narrow ideologies, to keep us connected to our families or craving more from our favorite brands, to give us quick dopamine hits or sustained serotonin levels. Its impacts are more far-reaching than we can possibly know.

The good news (and the bad news) is that any flaws in our newsfeed algorithms aren’t the fault of come computer; they’re flaws in the human design of the system. That also means humans have the power to fix it, to design feeds for connection instead of clickbait.

Tailored Retail Experience

In historic, small-town America, the general storekeepers knew the story of every person in town. They’d know how many people were in your family and their ages, what kind of food you bought each week, what clothing you liked to wear, and what kind of budget you had to work with. When you entered the store, they’d give a friendly greeting, help you find what you needed, and make recommendations for you—knowing exactly who you were and what you wanted.

Today’s retailers are aiming for the same kind of relationship, with the help of big data.

Whether you’re shopping online or using a store’s rewards card, the company is keeping tabs on what you buy. This data combined with machine learning helps them make personalized recommendations for you, and extend special offers that you’re likely to find helpful as well. Offers can vary from providing a coupon on your mobile phone when you’re near a store, including you on a targeted email campaign when they know it’s for a product you’ll probably like, or even create a special offer for your based on your online behavior on their website.

In the aggregate, they can use purchasing data plus info from social media, call centers, and product reviews to figure out what customers love, what they find frustrating, and how to make the overall experience better over time.

And on the back-end, companies can use data to improve inventory management, transportation routes, and negotiations with their supplier. The ultimate goal is a streamlined experience from manufacturer to storefront, from online to in-store, from advertising to purchasing, to make sure the products you want are where you want them, when you want them.

Target Your Career

If you’re a new or aspiring data scientist thinking about the next stage of your career advancement, learning more about audience targeting using big data, artificial intelligence, and machine learning can help make you more marketable in the workforce. From the advertising industry to politics to news media, the discipline of data science is helping people cut through the noise and connect to the information they need.