Five Ways to Mine Bank Data for Marketing Gold

Banks sit on a wealth of customer credit, debit and online banking transaction data, and they have a growing number of options for mining that data to determine what products customers want to buy and when.

A number of vendors of transaction marketing software are helping banks sort through their customer data using sophisticated analysis engines that look at sometimes trillions of bytes of information. Each vendor has its own spin behind the scenes, with some focusing more on the merchants through merchant-funded reward programs, others on the banks, and some trying to serve both the merchants and the banks.

Regardless of the target audience, all the vendors examine previous transactions to spot patterns and match consumers with an offer. Increasingly, they are applying layers of analysis to attempt to predict consumer behavior, such as the likelihood consumers will actually buy something and how much they are likely to pay.

Typically these companies make their assessments based on 15 months of data, but that is improving. Some are getting adept at making judgments with as little as three months of data and as few as a couple hundred transactions.

"We have to leverage our transaction data, so when we mail or email, we present targeted offers based on the member profile," says Jim Craig, vice president of marketing, 1st Advantage Federal Credit Union of Yorktown, Va.

The credit union began using a product from Micronotes Inc. of Cambridge, Mass., to cross-sells bank products to consumers at the conclusion of their online banking sessions.

"Only recently have both the analytic technology and the ability to handle very large data sets [converged] in a cost-effective fashion," says Steve Ledford, a partner at Novantas LLC of New York.

Geezeo Inc. of Tolland, Conn., a provider of a white-label online personal financial management product, offers a cross-selling service that helps banks and credit unions churn through their transactional data to identify other product offers for their customers. Geezeo's platform looks primarily at data provided by its customer financial institutions, but because its PFM tool also aggregates external customer data using Fiserv Inc.'s CashEdge, it has a larger glimpse of customers' financial lives.

It can help financial institutions do targeted campaigns, using an analysis engine that spots transactions by keywords such as "boat" or "home."

"The advantage of the PFM tool," which is built to include data from accounts with other companies, "is that financial institutions not only [see customer] transactions, but [external] transactions and financial products," says Peter Glyman, Geezeo's co-founder and president.

In the merchant-funded rewards category, Cardlytics Inc. and Truaxis Inc. use big data sets.

Both help banks encourage spending on their cards by using consumers' transaction history to present merchants' offers.

Cardlytics' software, like Geezeo's, sits in part on the bank's core processing system, sponging up transaction data. This data gets stripped of personally identifiable information before it is shipped to Cardlytics' data center, where it is mined each day for attributes that match criteria set by merchants.

"When we see a transaction that meets the criteria of an offer we presented, we update our software to let [the consumer] know they have earned a reward," says Lynne Laube, president and co-founder of Cardlytics.

By contrast, Truaxis, the Redwood City, Calif., company that until recently was named BillShrink, mines anonymized data across its network of bank customers.

"We have to deal with trillions of data points to triangulate the right information," says Schwark Satyavolu, chief executive and co-founder of Truaxis. He says the analysis engine tries to minimize the number of deals a customer gets, and that the system learns from consumer responses, offering the most deals only to the customers that more actively redeem them.

"The entire predictive analytics engine can be leveraged by the bank, along with detailed reporting and adaptive targeting that it enables," Satyavolu says.

FreeMonee of San Mateo, Calif., is different from other vendors because its merchant clients provide cash gifts, not coupons. The funds can only be spent from the bank client's payment card.

Gifts of this nature are redeemed at a 5% to 15% rate, whereas coupons are redeemed at less than 1%, says Jim Taschetta, chief marketing officer for FreeMonee.

To make that system work, FreeMonee has to vet the likelihood the customer will actually visit a store and make a purchase for more than the gift's value. So its analysis engine looks at dozens of variables to help it gauge whether the customer will actually buy. It then tries to craft an incentive for as little money as possible. The system then tracks customer redemptions.

FreeMonee's approach, which it calls collaborative filtering, relies on technology similar to Amazon.com Inc.'s search engine, which tells customers what other people have bought based on their own searches.

It also examines a customer's distance from the store, previous spending according to categories and likelihood to shop at a particular store. It uses a score that ranks customer spending levels at stores.

"The variables we use to predict are frequency [of purchases], spend and relationships between different merchants," says Andy Laursen, vice president of development for FreeMonee.

"As redemption comes in, it allows us to build different behavioral models that will actually predict and converge quickly," Laursen says.

Their gift-redemption model has the potential to add incremental revenue of $20 to $50 per customer card per year, Taschetta says.

"The Googles and Yodlees of the world have the ability to disintermediate the banks, and they are basically saying to consumers, 'If you give me permission, I can get a hold of all your transaction data,' " Taschetta says. "This is saying to the banks, 'Be careful: If you don't begin to mine this data and use this data, somebody else will.' And they are."