Moneyball for Marketing

Some say it was the beards and team chemistry that led the Red Sox to win the World Series last month. But many baseball insiders argue that the real heroes were Bill James and performance scouting. 

James is the “founding father” of sabermetrics, the science of using statistics to search for objective knowledge about the game of baseball. Performance scouting involves using statistics to evaluate ballplayers based on what a ballplayer has done, not on what he might do or what he looks like. 

Billy Beane, the GM of the Oakland Athletics, used James’s ideas to rebuild his team with cast-offs and misfits that were mathematically predicted to do well. He saw great success. The exciting story is told in Moneyball, a book by Michael Lewis and a movie starring Brad Pitt. When Theo Epstein, a disciple of James, became GM of the Red Sox, he built the team into pennant winners using the same principles. The beards may have helped, but mathematics is at the heart of this baseball triumph.

Just as how Moneyball popularized the search for better predictors of athletic performance, some marketers are looking for the right metrics to predict customer behavior. And it’s only recently that ‘Moneyball for marketing’ has gotten some traction. The marketing analog is called data-driven marketing or mathematical marketing. Just as in baseball, there are new metrics. Also as in baseball, a stubborn rear guard of traditionalists still resist the new methods. This month’s column goes inside this debate and examines the modern metrics that enable individualized marketing.

What’s wrong with the old metrics

For years baseball players have been evaluated by traditional metrics, such as batting average (BA), home runs (HR), runs batted in (RBI), earned run average (ERA), and pitch velocity. The best hitters were those who won the “Triple Crown,” leading their league in BA, RBIs, and HRs. Marketing has its own traditional metrics and its own “Triple crown”—recency, frequency, and monetary value, (RFM)—which has been in use for decades to evaluate customers.

Like the baseball metrics, RFM works to a certain extent, but it has serious flaws. It’s not capable of discerning trends in a customer’s behavior. With no product information in the metrics, RFM is worthless for measuring, let alone predicting, cross-sell. Logistic regression doesn’t scale well for large product sets. Marketers searching for objective customer behavior knowledge and ways to predict purchasing patterns have been forced (like Bill James) to look for new metrics.

The new math of marketing

In baseball, the sabermetricians invented metrics like “on base percentage plus slugging” (OPS), defense-independent pitching statistics, and a new value measurement called “value over replacement player” (VORP). The traditionalists cringed, but teams choosing players by the new metrics are winning championships.

Creative marketers are taking the same approach. New customer value measurements not only include recency and revenue, but they also include changes in inter-order wait times and breadth of purchasing. For example, “purchase delay” measures how many purchases a customer has missed from their usual pattern, a key indicator of deteriorating behavior. “Risk score” measures the likelihood of an individual customer not making another purchase before becoming inactive. It’s the best quantitative measure of customer loyalty and key to customer retention. 

Perhaps the most important of the new metrics is “purchase propensities,” which quantitatively identifies upsell and cross-sell products to offer to individual customers. These purchase propensities have finally cracked the cross-sell puzzle and are delivering, almost unbelievably, double-digit response rates for cross-sell.

The common threads uniting the new metrics are the following: They’re all based on what customers have already done, and they’re all calculated at the individual customer level. They’re a fresh attempt to capture objective knowledge about customer behavior. 

Obstacles and possibilities

In baseball the obstacles to this new math came from the crusty, tobacco-chewing scouts and managers who trusted their gut and their eyes more than the sabermetricians and their computers. In the marketing world, inertia and innumeracy stand in the way of progress. New metrics and tools are here. Automated, scalable, off-the-shelf software based on the new metrics is finally available.

What’s driving the adoption of the new methodology is the growing realization that customer-centric marketing based on these new metrics produces better results. Relevance raises response rates, and relevance comes from the new purchase propensities and the new measures of a customer’s state. A purchase isn’t needed for a triggered communication; it can now be based on a changing risk score.

The idea that this fresh look at predicting customer behavior leads to greatly improved ROI is the new reality for marketers. Sticking to established ways may be comfortable, but it doesn’t win championships. It’s not the beards that win games–and sales; it’s using what individual customers have actually done to understand what they’re going to do in the future.