PNC Cashes in on Data Cleansing

 
 
By David F. Carr  |  Posted 2004-03-01
 
 
 

PNC Cashes in on Data Cleansing


PNC Banks information-systems team wasnt thinking about compliance with the Patriot Act when it started a cleanup of customer files in 2000. The Sept. 11, 2001, terrorist attacks and the federal legislation that quickly followed hadnt happened yet. The new regulatory climate would ultimately demand that PNC be able to prove it knew its clientele well enough to sort the legitimate customers from illegitimate ones—and perhaps even identify potential terrorists.

The project that started in 2000 was designed to give PNC a better understanding of its customers so it could market to and serve them better. The bank wanted to cleanse its customer records of incorrect addresses and eliminate confusion about multiple accounts opened by the same individual under slightly different names.

That experience is turning out to be valuable in the face of increasingly stringent anti-money-laundering regulations, aimed not just at terrorists but also at drug kingpins and foreign leaders hiding money in U.S. banks. These rules make it necessary to match names against a variety of aliases and alternate spellings (particularly Arab names such as "Khaddafi," "Gaddafi," "Qadhafi" and so on), which is a variation on the same issue PNC learned to address through its data-quality project.

Check out eWEEK.coms Database Center at http://database.eweek.com for more database news, views and analysis.

The U.S. Office of Foreign Assets Control (OFAC) publishes "watch lists" of individuals and countries with whom U.S. companies are banned from doing business—thats not new. But the Patriot Act threatened stricter penalties, including fines of up to $10 million and prison sentences of up to 30 years for executives judged to have been criminally negligent.

Banks that had never cleaned up and consolidated customer data are being forced to do so as part of compliance with the new regulations, says Richard De Lotto, principal analyst with the financial-services practice at Gartner Inc. "Suddenly, you start looking at 30-year prison sentences, and it focuses your budget wonderfully," he says.

PNC Bank is the largest business unit of Pittsburgh-based PNC Financial Services Group, a diversified financial-services company with assets of more than $68 billion. The bank provides community-banking services, as well as wholesale banking services, to 2 million households and 190,000 businesses in Pennsylvania, New Jersey, Delaware, Ohio, Kentucky and Indiana. Although it has grown partly through acquisitions—most recently, the $649-million purchase of New Jerseys United National Bancorp completed in January—PNC has a strict policy of combining systems "on Day One," says George Kunkel, the banks vice president of management-information systems.

While many of PNCs competitors have found it more economical to let its acquisitions operate discrete systems in parallel, rather than integrating them, PNCs approach "shows that they care about the data, they care about their customers, and theyre trying to return profits to their owners," De Lotto says. Compared with banks that merely clean up data in response to the regulatory threat, PNC is ahead of the game, he says. "Thats one of the places where people screw up because they think data cleansing is not sexy."

Next Page: How data scrubbing affects customer relationships positively.

Scrub


-A-Dub for Better CRM">

Scrub-A-Dub for Better CRM

Kunkel agrees that, because the results are often indirect, "its hard to sell people on data quality." The direct effect of eliminating address-data errors, for example, is that PNC drove the number of returned mailings down 98 percent from where it stood in the first quarter of 2000, when almost 45,000 checking- and savings-account statements and other mailings came bouncing back. In the three months ending June 2003, PNC saw only about 1,300 pieces of returned mail. PNC calculates that it spends about 9 cents per piece processing returned mail (just the handling labor, not the wasted postage), which means its spending about $16,000 less annually than it was a couple of years ago. Not a big number for a bank that had revenues of $3.8 billion last year.

The real significance is that thousands of PNC customers arent left wondering where their statements are, and the bank isnt missing an opportunity to promote other services to them. PNC points to gains such as a 14 percent increase in ATM and debit-card fees and a 15 percent increase in home-equity loans as evidence it is, in fact, broadening its customer relationships.

Better communication with customers may be a factor in PNCs improved customer-retention rate, which has climbed from 91 percent to 94 percent since 2000. Thats impressive, given that the banking industrys average retention for checking-account customers is about 85 percent, says Kathleen Khirallah, a Tower Group analyst. Its generally much more profitable to retain customers than to recruit new ones, she says.

At PNC, data-quality efforts center on the banks so-called Customer Information File (CIF), a 21-year-old mainframe application. CIF stores data in the Information Management System, a hierarchical database from IBM. PNC also has invested in Siebel Systems customer-relationship-management software for its branches and its Web site, but CIF remains the banks official repository of customer data. So when a branch employee processes an address change in Siebel, that change is recorded in both Siebels relational database and in CIF. Some other changes, such as new account openings, are synchronized nightly.

Although PNC is evaluating the possibility of migrating all customer information to Siebel, for the time being Kunkel remains focused on refining CIF.

In 2000, PNC took two big steps toward eliminating errors in its customer files. In May of that year, the bank subscribed to a Pitney Bowes service that provides change-of-address data obtained from the U.S. Postal Service. That lets PNC bring its address records up to date and keep them current. A couple of months later, the bank began implementation of Trillium Softwares data-cleansing offering.

At its most basic level, data-quality technology is aimed at eliminating errors introduced by sloppy data entry. PNC has also invested in a type of data analysis known as "householding," aimed at understanding the relationships between account holders. For example, if "Elizabeth," "Beth" and "Liz" are all listed at the same address, are they the same person? Or is "Liz" the daughter of a woman listed on one account as "Elizabeth" and another as "Beth"?

Customers tend to be annoyed when their bank fails to make what ought to be obvious connections, particularly when the bank treats multiple accounts for the same person as if they belonged to different people. "As a bank, youre supposed to be good at accounting and other forms of bookkeeping," Gartners De Lotto says.

PNC previously used software from Brightware, now a division of Firepond, to improve the editing of customer records but wanted a system that would require less manual intervention. About two years before it started working with Trillium, PNC purchased and started to implement software from another vendor—one Kunkel says he is legally not allowed to name. The bulk of that implementation relied on the vendors employees or consultants, something PNC was eager to avoid in the future. With the implementation of the Trillium software and subsequent refinements, PNC project manager Greg Ford and his team quickly became self-sufficient.

"I believe we never would have got there with the other vendor. We would still be reliant on them," Kunkel says.

Trillium detects if an address doesnt exist, even though it may be formatted properly, by checking against a database of valid addresses. PNC immediately saw a big cut in returned mail.

One of the best ways to improve data quality is to get it right the first time. So Ford now works with the Siebel project team to integrate CIF interactive functions, such as address verification, into the graphical user interface deployed in bank branches. That will mean, for example, that when a customer comes into a bank branch to open a new account, the branch employee waiting on that customer will be able to detect potential problems with an address right away and correct them on the spot. So if that person lives on "1000 Maine Street" but it was entered into the new account screen as "Main Street," the address-verification system will be able to pop up an alert if no such address exists in that ZIP Code.

Today, address and watch-list matching exist as back-office processes that PNC applies to all new accounts.

Ford says he has learned its important to instruct the software on when not to mess with data that it would otherwise "fix." For example, there was the time the system kept automatically changing a customers address from "Apartment 10" to "Trailer 10," to match U.S. Postal Service records. But the change produced a swift complaint from the resident of that address.

"He turned out to be a senior vice president for some company, and he did not want anyone to know that his address was a trailer park," says Deborah R. Smith, vice president for technology delivery at PNC. While the system has to be flexible enough to make exceptions that respect a customers preference, she says, " we dont want to take away the good that this does."

Next Page: Matching the watch lists.

Matching the Watch Lists


Matching the Watch Lists

When the Patriot Act came along and OFAC rules were tightened, some of Trilliums other customers talked the vendor into providing a subscription service that would help them comply, and PNC helped define the requirements. Now Trillium digests the unstructured files delivered by the government and translates them into a format its software can understand.

Compliance with identity-verification and -matching regulations will cost PNC just over $1 million a year, Kunkel says, most of which is for back-office labor to investigate possible matches.

After several delays, the final rules for customer identification developed by the U.S. Treasury Departments Financial Crimes Enforcement Network and seven other regulatory bodies were published in April 2003, giving banks and other financial institutions an Oct. 1, 2003, deadline for implementation.

In addition to writing batch routines to search CIF for suspect records, Fords developers created an interactive application for researching those records and determining whether there really is a high-probability match that must be reported to the government. They also created a Patriot Act compliance solution for PNC to sell to its mutual-fund service-bureau customers.

Smith says household-account matching is also important for compliance with the Federal Trade Commissions Do Not Call List, which allows individuals to specify they do not want to receive telemarketing calls. The law doesnt care that a customer who makes that request may be listed under slightly different names for different accounts. "If we call them because we havent marked one record, we have a problem," she says. The federal law provides for fines of $11,000 per instance and some state laws are even harsher.

As far as the PNC folks know, none of the possible watch-list matches passed on to government investigators has turned up a genuine terrorist. "In a lot of cases, we never know," Kunkel says. "The government doesnt want us to know." However, one result of combing through customer files is that PNC turned up some 800 accounts whose owners identities could not be verified. That means PNC is now required to close them.

Whats more important for PNC is that the bank protect itself against the embarrassment of being found to harbor terrorist funds or of being found by a government audit to have failed to make a good-faith effort at preventing that sort of thing. "I think were in pretty good shape," Kunkel says.

PNC Financial Base Case


PNC Financial Base Case

Company: PNC Financial Services Group

Headquarters: 249 Fifth Avenue, Pittsburgh, PA 15222

Phone: (412) 762-2000

Business: PNC Bank is the largest business unit of the company, which also provides asset-management and mutual-fund processing services.

Chief Information Officer: Timothy Shack

Financials in 2003: Total assets, $62.8 billion; earnings of $274 million on total revenue of $5.3 billion. Revenue from banking operations dipped to $3.8 billion, down from $4.0 billion in 2002.

Challenges: Eliminate inaccuracies in customer data, create a more rounded view of each customer and drive up customer retention. Leverage this data to improve the banks response to laws involving customer-identity verification and watch-list matching.

Baseline Goals:

  • Grow banking business 5%-7% per year.
  • Grow overall financial-services business 10%-12% per year.
  • Use better knowledge of customers to improve retention from a rate of 91% in 2000 to 94% in 2003.
  • Reduce the volume of returned mail by 50%-65%.

Rocket Fuel