The mathematics of marketing, Part Two: Machine-learning, modeling, and Moneyball

There’s a memorable scene in the movie Moneyball where Brad Pitt—playing the role of Oakland Athletics GM Billy Beane—tosses out two hundred years of baseball intuition in favor of mathematical fact. As the film progresses, it makes a very interesting point: that one of the most valuable players on one of history’s most successful baseball teams may not have been a pitcher or a hitter, but a statistician. By letting data, and not gut feeling, drive his decision making, Beane and his team almost went to the World Series against incredible odds.

What does Moneyball have to do with marketing? More than you might think. Marketers can benefit tremendously by looking beyond business as usual and looking to the mathematics of data science to drive their campaign decisioning. Now, I know what you’re thinking: more data science means more data scientists, which means more delays. But that’s the old marketing model. The new decisioning model uses advanced machine learning to automate the analytical process and uncover valuable data insights and hidden content affinities in minutes, not weeks or months.

Machine learning does the tedious work that often falls to data scientists: identifying data correlations, discarding those correlations that lead to dead ends and building likelihood models based on various attributes. Using automation, marketers might be able to score their customers based on hundreds of different attributes rather than just a handful of demographic data points. Automation isn’t necessarily doing anything different than what data scientists are doing today, but it’s doing it faster while taking potentially problematic factors like intuition out of the equation.

Do you need data science? Yes!

Do you really need data science? Yes, you do, because segmentation alone is far from an exact science. Segmentation will tell you what your customers have in common—e.g., age, household income, basic buying behavior—but it won’t tell you why they’re more likely to buy a product or consume a piece of content. For that, you’ll need to cross-reference a wide variety of different content affinities and behavioral patterns. Marketers are understandably averse to engaging data scientists for fear of losing control of the campaign process while simultaneously slowing it down. Machine learning delivers the same (or better) results without the wait, which in turn enables marketers to do some exciting things with their data, such as:

· Populate recommendation templates with unique, personalized content;

· Segment customers into much finer groups based on specific content affinities (e.g., customers who like sports movies and movies that star Brad Pitt);

· Define what the next best experience would look like for various segments (e.g., a personalized email with recommendations, a coupon offer through social media, etc.);

· Identify what action/event should trigger an experience (e.g., What should happen after a customer buys a product from you? What happens when they haven’t purchased from you in six months?);

· Create lookalike models to acquire new customers and drive more revenue from existing customers through cross-sell/upsell offers.

Unlike Brad Pitt in Moneyball, you probably have the right team in place already. What you really need is the data science that can tell you how to create and position offers that resonate with your audience. Segmentation is a superficial solution to the problem: it may give you the illusion of fielding the best possible campaign, but the results speak for themselves. Data science delivers a deeper bench of data insights, although it can come at a high cost when done manually. By automating the data decisioning piece and orchestrating it so that customer experiences flow seamlessly through the right channels, marketing teams ensure that their audience is getting the best possible experience. And, for marketers everywhere, that’s the name of the game.