The Science of Persuasion

Ken Strasma pioneered a predictive-analytics methodology that transformed presidential elections. Now he’s applying those results to marketing, using data to map how people make decisions.

Ken Strasma, HaystaqDNA, CEO

“People almost invariably arrive at their beliefs not on the basis of proof but on the basis of what they find attractive.”  Blaise Pascal, The Art of Persuasion

You may be convinced that you know your own mind. Ken Strasma knows it, too. You may be an individual, a unique snowflake, an inscrutable cypher. No one can force you to change your opinions. But Strasma can figure out how likely you may be to do so. In fact, he’s betting his whole business model on it.

The philosopher Blaise Pascal, writing in the seventeenth century, understood even then that you can’t convince people to change. You can, however, learn the “hows,” “whys,” and “whats” of their attraction, and speak accordingly—especially with technology and data on your side.

Predictive Analytics

Predictive analytics (PA), a branch of data science used to analyze population data and make predictions about audience behavior, is becoming an increasingly pervasive tool with which to accomplish this audacious goal. The practice originally emerged from the credit-card industry in the late 1990s as a means for scoring and targeting potential credit-card users, but with increased computing power and data accessibility over the past decades, it has since been developed as a leading tool for actuarial science, political campaigns, and now, marketing and advertising.

Strasma is leading the development charge as cofounder and CEO of HaystaqDNA, a Washington, DC-based predictive-analytics firm founded in 2012. It is the sister company of Strasma’s microtargeting and data analysis firm, Strategic Telemetry, which popularized the use of PA for political campaigns in support of Barack Obama’s winning 2008 presidential run. And now that PA is part of what Strasma calls the “tool chest” for almost every major political campaign, he’s set his sights on the corporate realm through his work at HaystaqDNA.

A political campaign pretty much comes down to persuading people who are undecided, and turning out supporters. That’s the exact same thing that applies to pretty much any commercial endeavor.”

“We got a lot of interest from the commercial world after President Obama’s election,” Strasma says. “There are major parallels between the types of strategic thinking that go into selecting targets with PA models in politics and those selected for commercial campaigns. We’ve been able to apply that strategy and technology with great success in the commercial world.” Some of HaystaqDNA’s clients over the past few years have included HBO, the NFL, CBS, AMC, Audi, and even the New York Police Department.

Political Roots

Though PA is a relatively new method of targeting and analysis, Strasma’s approach—which takes its cues from the business of political campaigns—is informed by his hands-on political experience. He began his career in 1986, running a successful statewide campaign that led to his service as the director of state legislative caucuses in Wisconsin and Minnesota. He later headed to DC in 1997, where he became the research director for the National Committee for an Effective Congress. In this role, he produced geographic targeting for various Democratic candidates.

After founding Strategic Telemetry in 2003, Strasma went on to serve as the national targeting director for John Kerry’s 2004 presidential campaign, which would also be an early turning point for PA and microtargeting technology (direct-marketing data mining for predictive market segmentation) in political campaigns.

“We were able to use microtargeting technology to a great advantage for Kerry in the 2004 Iowa caucuses, a perfect petri dish for this type of technology,” Strasma explains. There was a small electorate that, at the time, had already been surveyed by dozens of different candidates for the campaign process. This meant it would be difficult for Strasma and others on Kerry’s team to reach these people—who were already burnt out by survey outreaches. Microtargeting allowed Strasma to mine extant data to create accurate predictions of how people would vote and how to best reach those people.

“We were able to use that technology to help John Kerry win the Iowa caucuses and go on to win the nomination,” Strasma says.

Though microtargeting proved to be valuable for the 2004 Iowa caucuses, at that point people were still generally skeptical about the reliability of PA models. “A predictive model is nothing but a prediction,” Strasma says. “But as people used it more, they found that in aggregate the models were remarkably successful.” Because PA involves powerful computing, complex algorithms, and accessibility to large amounts of data, general acceptance of the method—especially in the early 2000s—was slowgoing. People doubted its accuracy, but as they compared their extant predictive models to Strasma’s algorithms, the predictions proved themselves accurate.

The Future of PA

Over the next four years, computing speed increased, algorithms improved, and data became easier to mine, encouraging a wider embrace of PA and soon leading to an easy top-level buy-in for Obama’s 2008 campaign, which proved to be a turning point for PA and political campaigns.

“Computers were faster; storage was less expensive. We were able to build models faster than we could four years prior,” Strasma says of Obama’s 2008 campaign. “There was a cultural shift—people expected to have models as part of their campaign tool chest.”

While the cultural shift can be credited to increased acceptance and ubiquity of technology (for example, open-source software like Python and R programming with PA packages), computing, and data analysis for audience targeting, it also reflects a shift in overall marketing and persuasion philosophy. Rather than dispatching campaigns with an uninformed goal of engagement or conversion, Strasma’s approach to PA is that it’s a waste of time and resources to campaign—commercially or politically—to people who have already made up their minds. “Predictive analytics allows campaigns to engage undecided voters and buyers,” he says.

Microtargeting and PA allow campaigners to find that persuadable person, which is why the method works well beyond politics and into brand advertising and consumer goods. “While microtargeting allows you to focus on people who are legitimately more persuadable,” Strasma says, “it also gives you a powerful tool for persuading the people who we traditionally might have thought were not persuadable at all.”

He gives an example from the auto industry, where people make decisions with roughly the same regularity and information as they do with presidential elections. On one hand, persuasion can reach potential car buyers who are already in the market for a car; this is Advertising 101.

On the other hand, you might want to talk to someone who isn’t necessarily thinking of buying a car, but is deemed worthy of outreach based on their interests as processed through predictive models. “A political campaign pretty much comes down to persuading people who are undecided, and turning out supporters. That’s the exact same thing that applies to pretty much any commercial endeavor.”

While PA’s benefits are manifold across political and commercial campaigns, its ready application to both settings also reflects a current political sentiment—that the office of the president is just another product to be bought and sold. But this is precisely where the digital media milieu steps in to maintain checks and balances.

“There is always a danger of abuse when you have a technology that allows you to know what people think about a particular issue,” Strasma says. It is to PA’s advantage, he contends, that it emerged alongside the Internet reaching cultural ubiquity. “If an unscrupulous politician tailored their message to pretend to be on one side of an issue to attract one group, and then another side to attract another group, the Internet would catch them immediately.”

The Internet is not only a valuable social tool for keeping politicians in check—especially with targeted campaigns—but it is also the area PA is increasingly turning to for rich, compelling data for commercial and corporate campaigns. In politics, for example, mail and phone surveys used to be the best methods for gathering electorate data.

But now, as more households become difficult to reach through “traditional” one-direction marketing methods, the data users provide to the Internet and social media is an increasingly rich resource—though it isn’t always necessary. The real power is actually in the algorithm.

“There are good predictive algorithms available in open-source software, and the gap between those and stronger proprietary algorithms is getting smaller and smaller,” Strasma says. While the availability of data can give you an edge, he says, it’s not necessarily a decisive edge.

What is decisive, however, is the role of data for the future of PA. As Strasma turns his PA focus to marketing, he identifies three major areas where PA will continue to grow and develop over the next few years. The first is the continued breaking down of data silos. Strasma references Obama’s 2008 campaign and the fact that audience data from TV, direct mail, and fundraising was stored in silos and nonconversant. The same is true in the auto world, where dealer and finance data never talk. “We have been able to begin the process of integrating that data so that each group’s activities can inform the other’s,” Strasma says.

The second area for change will be in speed. It used to take several weeks for analysts to update PA models as they awaited information and updates, but now that smartphone and cloud technology is ubiquitous, new information can arrive from the field instantaneously to further inform the algorithms and their predictions. And the third area of change, data resources, is especially relevant for corporate and commercial campaign predictions, as sources become more numerous.

“Consumers are providing all kinds of information about themselves on social media,” Strasma says. “But the problem is that the information can be unstructured, so we’re working on ways to train artificial intelligence to translate online identities into real people.”

While PA also has applications in fields outside of campaigning and marketing—such as epidemiology and pharmaceuticals—the real development happens where the money is, which is why marketing holds the current lead. “In the last couple of years, there has been a lot of progress in advertising to the low-hanging fruit, such as people who like or follow a page,” Strasma says.

“But the industry is only now getting better at talking to people who haven’t engaged on that level yet. There’s definitely a lot of room to grow in the commercial and political worlds.”