Predictive analyctics help customer retention

May 26, 2003

BY MICHAEL KRAUSS

Wouldn't it be great to know what the future holds and how to prepare for it? The ancient Greeks had a way: Years ago I traveled to Delphi in Greece, a beautiful place with magnificent vistas. There, at the ruins of an ancient Greek temple, the tour guide explained how they did it. He said they'd feed intoxicating laurel leaves to a woman who was designated as their oracle. This unfortunate oracle would spout a few things in delirium. Then the priests of the temple would go off and interpret the meaning of the oracle's ravings, making them conform to their own intentions for the community.

Let's face it, in today's ROI-driven marketplace, soft, qualitative approaches to predicting future trends are too unreliable. We need fact-based, rational tools that can aid marketing managers with their decision-making. Plus, there's such an abundance of data thanks to new technology that it's often hard to know where to start. That's why I'm glad I know Jack Noonan. Noonan, president and CEO of Chicago-based SPSS Inc. (a developer of statistical software) is leading the movement toward "predictive analytics."

Sorting through the data

In a recent interview, Noonan cited a major telecommunications company that was awash in data. Like most phone companies, this one is constantly extending new offers to different market segments to try to attract and retain customers. Their competitors are doing the same thing. The result: customer churn, a major industry problem, as smart but disloyal customers shop for the best deal.

The cost of this churn to Noonan's client was in the range of $300 million annually. The savings, by adopting predictive analytic software and processes, "was about $100 million," he related. Not a bad deal.

"The focus of predictive marketing is to help marketers look into the future," Noonan says. "It's to help them use their data to understand current conditions and potential future events and apply that knowledge to marketing programs to change customer behaviors." And maximize profits, I might add.

Shared goals

"Marketing managers have always wanted to do a handful of things effectively with their customers," Noonan says. "You want to increase the number of customers. You want to keep the best ones longer. You want to sell customers more products and services. You want to improve the outcome of your servicing efforts (in, say, health care). You want to find bad customers and get rid of them."

Listening to Noonan made me a bit nostalgic. When I first used analytic software years ago, I wanted to do exactly what he'd described. I wanted to find new ways to segment my customers so I could customize programs to unique segments and up-sell them.

I remember it was a cumbersome process mainly geared toward quantitative market research experts. The software (which I had to program) enabled me to reduce large quantities of customer survey data into manageable segments of buyers. It wasn't user-friendly, and I never would have completed my project had it not been for an advanced analytics specialist from the market research firm I'd retained.

Moving into the enterprise

Today, SPSS still sells to the guru market research analyst, but Noonan's real aim is to make SPSS an invisible, behind-the-scenes, enterprisewide software solution that will integrate multiple kinds of data to enable better decision-making and improved future performance.

Noonan's holy grail is to take real customer transaction data, meld it with customer characteristic data and blend it with traditional customer attitude data to gain a powerful information cocktail that can be mixed by his software to yield scenarios that forecast the future. He wants to place the software that blends this data right at the fingertips of the marketing manager and make it as simple to use as going to the water fountain to take a sip.

By way of example

Consider a marketing manager in a major bank. Thanks to improvements in operational software, it's now fairly easy to understand all the activities across the bank for a particular customer. A good customer might have heavy ATM use, a checking account, a savings account, a home mortgage and two credit cards. They might be heavy transactors generating plenty of fees. At his fingertips, the marketing manager can see the value of the bank's relationship with the customer.

By buying characteristic data from an outside firm, the marketing manager can see that this customer has great demographics. They live in the right ZIP code to be high net worth individuals. Then through traditional telephone-based market research, we might learn that they are highly prone to switch banks with a modest incentive. We can learn about their loyalty early enough to act to retain them. Because we have their transaction data and characteristic data, we can gauge that it's financially prudent to invest in retaining them.

Knowing transaction history, customer characteristics and customer attitudes can help the bank craft more efficient and effective marketing retention programs in real time. The net result is lower-cost marketing programs, greater revenues and profits, and increased ROI.

Find the changes

Noonan seems genuinely excited by the opportunity for predicative analytics.

"If you're going to improve in the future, you have to change something today," Noonan says. "Now we have software that can help us identify what changes to make and help us improve our performance in the future."

Truer words were never spoken by the Oracle at Delphi.

Michael Krauss can be reached at michael.krauss@diamondcluster.com or news@ama.org.