Ken Strasma first gained notice in 2008, when his techniques for microtargeting voters helped land Barack Obama in the White House. Strasma did the same for Michael Bloomberg the following year, when he was running for his third term as New York’s mayor, and then again for Obama in the 2012 presidential election.
As CEO of Haystaq DNA, Strasma and his team use the same predictive analytics techniques to figure out what consumers really want from products and how to get the word out.
Strasma recently spoke to CMO.com about the basics of microtargeting, how marketers can use data-based models to gain new customers and cut customer churn, and how choosing soft drinks is similar to a presidential campaign, while other product categories are more like a local election.
CMO.com: Can you explain how microtargeting works?
Strasma: At its most basic, microtargeting or predictive analytics is about taking what you know about some people and using it to make predictions about everyone else. We use a lot of statistical algorithims and data sources, like Census demographics and commercial marketing data, in order to build accurate models.
The speed and accuracy of this type of modeling has increased dramatically over the last couple of years, but the basic concept remains the same: We’re using information we have about some people, [then] combining it with data about everyone, to build statistical models that predict how those other people would approach that behavior.
CMO.com: How has microtargeting evolved since the 2008 presidential campaign?
Strasma: We’ve been doing a lot more commercial work lately as people have seen the technology proved in a very public forum.
In terms of how the technology has evolved, there are two tracks: One is on the technical side, where there’s constant R&D, finding better statistical algorithims, and the computing technology is getting faster and faster all the time. Equally important has been the understanding from end users.
We saw that first in the political world. These are people who used predictive analytics in every campaign they’ve run, and they’ve come to accept it as part of their normal toolbox.
We’re seeing that a lot more in the commercial world, as people begin to adopt that, too. They are demanding not just a good CRM list of past customers, but want to know predictive models about who’s likely to be future customers. That sort of understanding and adoption has been as important a change as the improvement in the technology.
CMO.com: How have marketers adopted this discipline to choosing among many products, instead of just two presidential contenders?
Strasma: You can think of a presidential election as Coke vs. Pepsi, where everyone knows the contenders. Other commercial challenges, where there might be hundreds of similar products, it’s more like a small local race, where people don’t really know the candidates.
Our challenge is to define the people who would support our candidates if they really knew about them. Likewise, our challenge is to find the people who would support Product X if they know about it, so the company can spend marketing dollars raising awareness not across the board, but among the people who will then buy the product. That brings a great deal of efficiency.
In the political world, you win or lose based on if you get one more vote than the other person one day every four years. In the commercial world, it’s more of an ongoing process where, if you don’t hit your sales numbers one month, you can retool and try to do better the next month. So the commercial world has tended to be more iterative.
CMO.com: How important is to playing to the base, to use a political term?
Strasma: Of course, none of this happens in a vacuum. The competitors are always trying to steal away your customers.
One of the most powerful uses of modeling is identifying customer churn and finding the people most in danger of defecting to a competitor. One of the reasons that’s so powerful is it’s one of the areas where we have the most data.
We can model what features people are likely to value, so if you’re trying to keep someone with you or upsell them, you know what features to promote. You can also know what features they might not like, so you know what anchor points might drive someone away.
All that can be modeled so the customer service rep on their screen knows these are the next three products this individual is likely to care about and be concerned about. That can be very valuable and useful information.
CMO.com: How useful are digital media in generating analytics. Can there be information overload?
Strasma: There is a danger of getting swamped. It’s a big shift in the move from very structured survey data to unstructured data online. One of the ways the technology has evolved most in the last year or so is advances in natural language processing—to learn sentiment from tweets and Facebook posts, and also learning to match real-world identities to online identities. Those are two important things to making use of this data.
The online advertising world has gotten to the point where people are very good at retargeting ads with cookies. The real next frontier for us is modeling people who are going to be in the market before they exhibit that shopping behavior. Once they begin shopping, everyone is going to be serving ads to them. We want to be able to model who’s about to start shopping for the product and get that ad to them first, when we have an exclusive conversation with them.
CMO.com: You’re speaking at Brandworks University conference in May about breaking old habits and making new ones. How does that work?
Strasma: One of the key points is knowing the value of someone as a recurring customer. To switch from a competitor the first time, breaking that habit can be extremely valuable. We’ve seen that in politics where once you get someone to vote for a different political party for the first time, they’re much more likely to split their ticket in the future.
It’s the same way with brands: People may have a particular brand of car or soft drink they always go back to. Getting them to consider a different one for the first time and knowing when they’re going to be in the market is really key and something where modeling can help a lot.
CMO.com: How? You use data to find touch points and emotional triggers?
Strasma: Exactly. We combine two different types of models: one where we know when someone is likely to be in market—the time to talk to them when we might potentially make or break a habit. And also we know what they care about, what they might be dissatisfied with about the product they are in the habit of buying, and what they might value in the alternative we’re offering them. So then we can come to them at the right time with the right message.
CMO.com: Does creativity come into play to use the knowledge gained from targeting?
Strasma: Absolutely. I’d say predictive analytics is a both an art and a science—I’d probably say two-thirds art and one-third science. If you focus entirely on the data science aspect, you might make good models, but you won’t predict the kinds of things that people care about, and you might not be able to make the leap from being an interesting report to being an actionable part of a marketing strategy.
CMO.com: What’s the next evolutionary step?
Strasma: I would say social media data mining is probably the biggest untapped growth area. A very, very small fraction of the available public information is currently being analyzed.
People are going to be getting smarter and smarter about how to use that information. I’d see that as the biggest growth area over the next year.
Editor’s Note: Ken Strasma will speak at this year’s BrandWorks University annual conference May 13-15, explaining how data modeling can reveal habits people don't realize they have, and messaging and media strategies that can tap those behaviors before they happen. CMO.com is a media partner.