Bank of America Case Study: the Information Currency Advantage Felipe Carifio Jr
Total Page:16
File Type:pdf, Size:1020Kb
Bank of America Case Study: The Information Currency Advantage Felipe Carifio Jr. [I] and Mark Jahnke[2] NCR/Teradata Parallel Systems 100 N. Sepulveda Blvd. El Segundo, CA 90245 [l] [email protected] [2] [email protected] Abstract This paper’s main focus is the banking industry and the experience many banks have had This paper describes the external forces that since the 1980s using the Teradata F&DBMS to motivate financial institutions to collect, build data warehouses [Inm92]. These Teradata aggregate, analyze, and mine data so that it data warehouses have been used to re-engineer can be transformed into information, one of a banking processes [Ham901 for customer financial institution’s most valuable assets. information databases in order to create unified In this paper we refer to this strategic customer profiles for relationship banking (also information asset as “information currency.” known as “householding,” “cross-segment In general, we describe the state of banking marketing,” “target marketing,” “databased and the rapid global changes that affect marketing,” or “marketing to the segment of one.” financial institutions. We analyze how Bank The data is sourced from diverse geographical of America (BofA) created and employed its areas, different lines of business (for example, information currency using the TeradataTM checking, savings, auto, home, credit cards, and Relational Database Management System ATMs), and/or from various online transactional (Teradata RDBMS). The Teradata RDBMS systems (such as IBM’s IMS, DB2, or CICS, as manages a very large data warehouse (NCR well as systems from DEC, Bull, NCR, Unisys, Scalable Data Warehouse) for BofA using an Siemens, Tandem, Fujitsu, Hitachi, and many NCR WorldMarkTM 51OOM MPP (Massive others). Parallel Processing) platform [Wck93]. In [CK92] we described three categories of 1 .O Introduction data warehouses: precision, discovery, and cross- Financial institutions are drowning in a functional views. That paper provided a Sargasso sea of raw data pounded by waves of “discovery” banking example for an institution new applications and services. If money managers that used its information currency to detect are not careful, these waves may become a customer patterns. The bank data mined tsunami, dashing them against the jetty of information about customers, which resulted in financial ruin instead of providing an opportunity realization of the following factors and to surf safely to shore [Arm96]. subsequent decisions: 1. Retention of customers with three or more Permission to copy without fee all or part of this services was high. Consequently, the bank material is granted provided that the copies are not targeted sales of new services to these made or distributed for direct commercial customers and developed other retention advantage, the VLDB copyright notice and the title of the publication and its date appear, and notice is programs. given that copying is by permission of the Very 2. Identification of customers shopping for the Large Data Endowment. To copy otherwise or to highest interest rate. These customers were not republish requires a fee and/or special permission marketed to, resulting in a reduction in the from the Endowment. Proceedings of the Uth frequency of opening and closing accounts. VLDB Conference, New York City, 1998. 641 The remainder of this paper is organized as BofA decided to implement [Kov97] a corporate follows: Section 2.0 describes the current and marketing web site (intranet) using Teradata. BofA evolving state of banking. Section 3.0 provides a uses a 2-terabyte Teradata data warehouse running Bank of America (BofA) case study on on a nine-node (8 CPU) 5100M WorldMark server information currency. (Note: information with 1,700 users. The Teradata data warehouse currency is our terminology, not BofA’s). started in 1986 with about 25 users and 20 GB of data. By early 1998 it had over 1,700 users and 2 2.0 State of Banking terabytes of data. Over 40 OLTP systems feed the Global deregulation, consolidation, and Scalable Data WarehouseTM (SDW) containing 38 privatization-which reached U.S. shores in the million accounts. Data analysis includes: 1) 1980s and 1990s-allowed interstate banking and profitability scores, 2) Channel propensity scores, 3) permitted banks as well as insurance, mortgage, and product propensity scores, 4) behavioral scores. The mutual fund companies to enter each other’s Teradata SDW is fully integrated into the BofA Call markets. This set off a flurry of mergers, takeovers, Center (CTI). and acquisitions. The U.S. banking consolidation BofA has the largest data warehouse in the began in 1985, and additional massive future banking industry and maintains information going consolidation is expected [Bro96]. The U.S. still has back five years on 36 million customer accounts more than 10,000 banks [Ecol-961. The 1995 from 30 different operational OLTP systems, GATDWTO free-trade agreement increased including data on checking, savings, time deposits, competition worldwide. Bank privatization in ATM transactions, real estate, consumer loans, bank Europe raised $22.5 billion from 1985-1995 cards, and commercial loans. [McR96], while bank privatization in the developing BofA always had a lot of data, but the key to countries totaled $16 billion from 1990 and 1994 success is the database’s transformation into [Eco2-961. information currency for use in the following: In this increasingly competitive, deregulated, and l Target and cross-marketing global market, banks need to attain certain large l Acquisitions economies of scale in order to compete. BofA can l Relationship banking determine the merit of a potential acquisition by l Credit card tracking integrating the consenting candidate company’s l Customer intimacy master-file tapes into BofA’s investment analysis l Retail banking format. Not only can potential takeover and merger l Portfolio analysis candidates be identified swiftly using the Teradata l Credit risk management RDBMS, but also-once an institution is A bank manager successfully used the system to acquired-its customer information can be quickly detect that some customers were leaving because of integrated into the existing Teradata data warehouse. unhappiness with the bank’s fees. BofA was able to Technological advances-such as the Internet, offer a low-cost checking plan targeted to these ATM machines, automatic payroll deposits, banking customers, thus halting the departures. BofA is by phone, and home banking on PCs-are trying to create the kind of intimacy between itself dramatically altering the banking industry by and its customers that existed 20 years ago. lowering entrance barriers for newcomers and The challenge to information currency is to increasing competition from new directions. These provide greater access to detailed data for customer technological advances may even make vast segmentation, clustering, and retention analysis as networks of physical branch banks obsolete. well as allowing more sophisticated analysis of the 3.0 Bank of America: Case Study banks financial metrics. The benefits of meeting this challenge are 1) decreased loan defaults, 2) Bank of America (BofA), based in San increased customer retention, and 3) increased per- Francisco, is the third-largest banking institution in customer profitability. the U.S., with 1996 sales of $22 billion and profits Customers are placed on a continuum from most of $2.9 billioeoth up eight percent from the to least profitable. The goal is to retain profitable previous year (Fortune, 4128197). customers and deepen relationships, target products, 642 and convert high “potential” value customers to high Numerous traditional and non-traditional value. Data quality is checked by plotting external competitors (for example, GM, GE, Microsoft, and data regarding customers’ yearly income with the Charles Schwab) prowl the information currency bank’s internal data (BofA loan applications). waters. To avoid shrinkage of the customer base, Information fed to the SDW includes checking, BofA plans to use Cray supercomputers in savings, time deposits, investment products, conjunction with the Teradata RDBMS to identify commercial loans, small business loans, consumer more profitable sales opportunities before the loans, real estate loans, credit cards, branch channel competition identifies and devours them [Ver95]. transactions, AMOS transactions, telephone For example, the BofA sales team can now sculpt banking transactions, PC Banking transactions, detailed demographic views of select groups of wholesale product & customer MIS, and global customers and then tailor its offer for remortgage capital markets. External data includes Donnelley loans and other financial products. Also, BofA can Demographics, Dataquick R.E. Loans (all lenders), now data mine for targeted information (such as how Claritas PRISM codes, and Dunn & Bradstreet. many of the 6,000 Silicon Valley residents in a Loads can be daily, weekly or monthly. Data can particular sales district own Acura Legends and golf be categorized as account-centric, customer-centric, club memberships, or which Hispanic customers are product, ownership, sequencing, geography, potential first-time home buyers) [Hof95]. profitability, and “scoring”. Access types are: 1) BofA uses the Teradata RDBMS to analyze Browser Access, 2) Analyst workstation, 3) trends in its relationships with customers to Advanced analysis. Browser Access means standard determine which are likely to purchase