In 2011, the term “Moneyball” went from inside-baseball (literally) to mainstream. The movie starring Brad Pitt, based on the 2004 book by Michael Lewis, was a hit, and “Moneyball” quickly became a metaphor for data-inclined business strategy, even if the line from on-base percentage to quarterly earnings wasn’t crystal clear.

In baseball, though, the moneyball approach has trickled up, from the field to the ticket office, as evidenced by efforts such as the Milwaukee Brewers’ use of analytics to better identify its most loyal customers: season-ticket buyers.

Last week, Knowledge@Wharton recapped a presentation about the Brewers’ analytics program at the Wharton Customer Analytics Initiative conference. (The full 38-minute session can be viewed on the WCAI blog.) Matt Horton, senior manager of advanced analytics at Major League Baseball, and Diny Hurwitz, data analyst with the Brewers, shared their process for using past data about ticket buyers to predict which ones would be most likely to buy or renew their full or partial season-ticket packages.

You might not know where your most powerful member data is lurking until you go looking for it.

For membership pros at associations, their process is a fascinating example of the power of advanced analytics, and season-ticket sales is a much closer analogy to association memberships than any on-field performance stats.

The model Horton and Hurwitz built analyzes data from five sources: transactional data from previous seasons, ticket usage (i.e., which pre-sold tickets were actually used to attend games), email tracking data, third-party demographic data, and ticket-buyer survey responses.

They found a set of variables that were best linked to a customer’s likelihood to renew a ticket package, including:

• month of the customer’s last purchase in previous year
• customer’s age (older being better, plus a bonus if born between 1929 and 1956)
• length of residence
• number of games tickets used in previous year
• buyer type: single-game buyer, package buyer, or both
• email open rate
• response to a “How likely are you to renew?” survey question

And they’ve found several interesting relationships between some variables, as well. For instance, the customer’s survey response becomes less important as the number of game tickets used increases.

The details of the Brewers’ predictive model are, of course, unique to the baseball business context—and many of the variables their model is based on either don’t apply or don’t even have rough parallels in an association’s membership operation—but the process they followed is worth examining.

First, they’re using a large amount of data stored about their customers, tracked over several years. And they started with just transactional data and demographic data, adding the other inputs in ensuing years and iterations to improve the model. The analysis—the technical mechanics of which, I admit, are well outside my skill set but are shared in some detail beginning around 5:20 in the presentation video—sorted through about 1,000 potential variables to arrive at the best nine to use in the predictive model.

Perhaps most interesting about their model is that variety and specificity in the predictors used, which range from buying habits to demographics to engagement (tickets used) and focus on some narrow parameters within each. Its power, though, lies in how it can be used to improve upon whatever methods or intuition you might rely upon otherwise. As Horton and Hurwitz explained, they used the predictive model’s ranking of customers to adjust sales tactics to them accordingly. For instance, by identifying those most likely to renew, the Brewers’ sales team could work more efficiently, dedicating more calls and outreach to “on the fence” customers, knowing that the best customers might renew on their own or need just a small amount of prodding.

Associations that are getting into predictive analytics are hoping to find similar advantages. The American Counseling Association began such an effort in 2013, using analytics to parse a wide range of member characteristics, response rates, and buying habits to identify four clear segments of members—which it calls Traditionalists, Go-Getters, Starving Students, and Value-Driven—and then rework member-marketing materials accordingly.

The Society of Critical Care Medicine, meanwhile, instituted a “single database rule” several years ago to get a better grasp of all the information it had about its members. In 2011, that led to an effort using predictive analytics to identify likely future volunteer leaders based on their early engagement patterns. “In other words, how can we predict which people are going to have those high [engagement] scores later by their current activity?” David Martin, CAE, CEO and executive vice president of SCCM, said at the time.

Associations track and store a wide range of information about their members, but I have often heard sentiments about data collection that amount to “don’t collect it if you don’t plan to use it.” The Brewers example seems to suggest that more data to choose from is better and that you might not know where your most powerful data is lurking. You just need the analytical power to go looking for it.

How is your association trying to predict your members’ likelihood to renew? Have you adopted a “moneyball” approach to recruitment, retention, and engagement? Share your thoughts and experiences in the comments.

(Jeramey Jannene/Flickr)