Data Mining to Identify New Markets (1,176 words)

December 2000

multipage

Another interesting fact the DTCU learned is that it has almost 100-percent checking account penetration within a seven-minute drive of branch location. "Then it drops off dramatically. The real interesting thing here is that it's based on time, not distance. It shows that people want to be near a branch—whether they visit it or not. This is the reason the dot-coms are having so much trouble," asserts Thompson.

He adds, "Now we've taken this, overlayed it with Acxiom data and profiled with lifestyles, and Acxiom gives us back these clusters so we can decide where to build future branches."

As its member base grows, the Dallas Teachers Credit Union averages two new locations per year. More factors come into play than just the number of members when building a new branch. Sometimes the new branch needs to alleviate high traffic at an existing branch.

"We just opened a new branch in a suburb that is relatively close to a branch that was overloaded," Thompson says. "We wanted the drainoff."

One-To-One Marketing

By mining the data in its warehouse, DTCU was also able to learn some interesting things about its existing customers, such as the predictability of bankruptcies. "We looked at the data warehouse to find what about bankruptcies was predictive and then could track those individuals more closely," Thompson explains. This is important, since part of DTCU's charter as a credit union is to serve the community.

If it started to look like a problem was developing for a certain member, DTCU could offer individual counseling services in-house, refer them to an outside agency or offer alternate payment plans.

"Another way we're using the data is running a cross-sell model with part of the IBM solution," adds Thompson. Through EZMart, DTCU has been able to run segmentation models on its customer base to see which segments should receive specific promotions.

Predictive modeling also helps determine in what sequence people get certain offers, based on answers to questions such as, "What products do people get initially"—say a checking account, an auto loan or a credit card? Adds Thompson, "We have 107 scenarios in test right now. It's a very complicated algorithm, but with great potential."

To help it in cross-selling certificates of deposit, individual retirement accounts and other investment vehicles, DTCU is using Unica Corp.'s Affinium predictive modeling suite to determine which product information and promotions are most likely to meet members' needs.

The next phase of the program implementation will allow DTCU to offer truly customized one-to-one service to its members. "IBM's e-commerce folks will be in soon to talk to our call center reps about how to use this system," explains Thompson. "Basically, what will be able to happen is when a customer calls in, and his or her number and information pops up on screen, a cross-sell screen with relevant product offers will pop up, too."

As an example, Thompson says, "If a client's auto loan is nearly paid off, we'll know that he had a Chevy Suburban and could work out a promotional arrangement with a local Chevy dealer."

In the future, Dallas Teachers Credit Union also plans to use demographics and traffic patterns to determine the placement of billboards and ATMs.

The Price of Results

None of this technology comes cheap, of course. Thompson admits it all would have been out of reach for a smaller financial institution like DTCU without exceptional circumstances: The organization had agreed to be the test site for the IBM pilot program.

Was it worth being the guinea pigs? Thompson says certainly it was, though there are always drawbacks to being first. "IBM gets to template from this and use it for future roll outs. So other companies will benefit from all the time and effort we put into learning the systems," he points out. "We benefited because our financial commitment was not as great as you might expect. We could not have afforded it. And IBM provided a lot of support." Thompson jokes, "We're on the larger side for a credit union, but we're no Bank of America. Where they might have 600 programmers they could pull from, I have a total staff of 15 with two programmers. Here we all have our normal day-to-day jobs of running the financial institution in addition to working out the kinks in this new system."