BMO Financial Group (formerly the Bank of Montreal) uses a central data warehouse with application-specific data marts to get that 360-degree view of its customers, and to identify which are most profitable, most likely to respond to an offer, or in danger of switching to another bank. The system won the Data Warehouse Institute's best-practice award for use of a data warehouse for customer-relationship management.
BMO's data warehouse, assembled five years ago, uses IBM's DB2 Universal Database software running on an IBM RS/6000 SP server. The system now has 5-to-6 terabytes of customer data (8 terabytes by next year) pulled from 42 operational systems, including retail, investment, commercial, and wealth-management applications. Software from Ab Initio Software Corp. handles the data-transformation chores.
From the data warehouse, the data is moved into data marts for specific analytical tasks. One is used to calculate the profitability of individual customers. "That was one of the most important data marts we ever built," says Ted Mendes, database management services director. "You need that as a base for CRM."
BMO uses software from SAS Institute Inc. to build models that groups customers into 12 categories, such as "empty-nesters" and "young families with children," and analyzes data for customer retention, profitability, and propensity to buy additional services. IBM software is used to attach scores to individual names, such as a customer's likelihood to accept a home-loan offer or leave for another bank. These scores are re-run each month, Mendes says.
The scores are then fed into an application from Unica Corp. that develops recommendations for each customer, such as offering them a new credit card or home loan. Those, in turn, are loaded into applications from Siebel Systems Inc. used by the bank's tellers, call-center works, and other customer-facing employees and systems. The data warehouse is updated to reflect customer responses to the offers. Business-intelligence software from MicroStrategy Inc. is used to analyze customer scores and offers to see what's effective and what isn't.
Mendes' team is developing triggers for the data-warehouse system that can automatically respond to customer actions. A customer making a large deposit that boosts his account 20% or 30% above his regular balance might be offered a certificate of deposit or an investment opportunity, for example.
AT&T Wireless Services operates two data warehouses that together hold about 11 terabytes of customer-billing and call-detail records. The systems are used by some 4,000 employees and managers in customer service, finance, and sales and marketing for everything from fraud detection to calculating sales commissions to reporting the division's financial results.
With so much riding on the accuracy of the data, maintaining data integrity is critical. The processes and policies AT&T Wireless relies on to maintain data integrity earned the division The Data Warehouse Institute's award for data stewardship and data quality.
The data warehouses are an older Wisdom (wireless information system, decision enterprise management), which will be retired next year, and a new-generation EDW (enterprise data warehouse) that was built last year. Both are based on NCR Teradata hardware and software. Users extract data from the warehouse using reporting software from Business Objects SA and Web-based applications developed using Macromedia Inc.'s ColdFusion.
AT&T Wireless' data-integrity program, begun in 1999, emphasizes data stewardship and gives the users of the data warehouse ownership of the data within. Designated representatives in marketing, for example, "own" data that pertains to distribution channels, which makes them responsible for the reference tables that define that data. Finance likewise is responsible for terms such as customer churn.
"Users have to be able to make the best decisions with the best-quality data," says Deanne Larson, data warehouse program-management director. The goal of the data-integrity program is to assure users that the data is meaningful, useful, and provides value to the business, she says.
A rigorous change-control process ensures that changes in data definitions don't negatively affect other data. Proposed changes are simulated using test scripts developed with Teradata utilities to gauge their impact. Data-warehouse managers also conduct extensive audits in conjunction with users when adding a new data source to the system to be sure it will meet data-integrity standards and user needs. And AT&T Wireless has well-defined procedures for users to submit questions and feedback to the data-warehouse management team and processes for investigating those issues.