Fleet Bank Launches Campaign Based on Data Warehouse
The mailings, sent out this month to Fleet's 10 million individual and institutional customers, market the company's insurance, discount brokerage, equipment leasing and investment management services. Executives hope the new system will provide better, more reliable and more timely customer and transaction data.
Before using this system, Fleet used a closed-loop system from Harte Hanks, Billerica, MA, which produced a marketing database with data from four external sources and included name and address standardization. Fleet, however, thought the process didn't offer any historical information, was time-consuming and found the data was out of date because it took two or three weeks to be loaded.
"Harte Hanks would offer us information one month at a time," said Bill Lonsdale, vice president of customer data management and analysis group at Fleet Bank. "While some of the information had been maintained over time, it did not have a great richness of history."
The new enterprisewide, open data warehouse, implemented in July, not only gives Fleet a more comprehensive view of its customers, but it's more robust, which allows Fleet to hold a great deal of data over a long period of time. Eventually, this will let the bank analyze its customers from a historical perspective.
Fleet, the 10th largest commercial bank in the United States, can keep daily customer account, transaction and demographic information from 34 internal and external data sources, including its ATM, branch, deposit, loan, investment and telephone banking systems, and will hold summary tables of this information for three years. The new inhouse system also gives it more control over its own environment.
"We think there are significant benefits," Lonsdale said. "Since Fleet is now directly involved in bringing together names and addresses at the customer level to create household aggregation, if we need to change the way that we do matching, we can immediately come up with a hypothesis of how the new matching should be, test it and then implement any changes."
For example, Fleet can select a group of customers or prospects and create a campaign to work against this group. It can execute the campaign, specifically analyze the results, test the output of the results analysis and use that as a guide to see how the campaign should be run next time.
The new system gives Fleet "the opportunity to do continuous learning, where we can analyze campaign one and feed that into campaign two," Lonsdale said.
However, he said it's unfair to compare the new system with Harte Hanks' system, since the two are totally different and represent a fundamental change in philosophies.
The data warehouse is from Informix Corp., Menlo Park, CA, and includes a decision support option that lets it analyze large amounts of queries quickly. In addition, Fleet is running a marketing data mart that employs data mining tools from SAS Institute Inc., Cary, NC; a campaign management tool from Exchange Applications Inc., Boston.; and a query generator from Cognos, Burlington, MA.
The data mart is an important part of Fleet's data warehouse, Lonsdale said.
"Some enterprisewide data warehouse projects are really the infrastructure piece, but what we have done here is [implement both] the infrastructure piece and the marketing data mart -- which allows us the ability to generate well-thought-out marketing campaigns and allows us to analyze customer behavior, attributes and characteristics," he said.
The data warehouse collects incremental loads, adding data such as updates to customer profile, account and transaction information, along with new customer and prospect information. Then, it is automatically updated daily, weekly or monthly, depending on the data source.
Currently, the data warehouse and marketing data mart are in full production on a cluster of Sun Microsystems 6000 servers.
The data warehouse -- loaded with 350 gigabytes of information -- is expected to grow to 4 terabytes. While it supports 125 users, Fleet is planning to have 1,000 employees using the system by the end of 1999.
As for the future, Lonsdale said the company is planning to add more data and capabilities to the warehouse and he plans to do more longitudinal, or "trigger-based marketing, where we will be generating marketing activity based on what we captured in our data warehouse," he said. "Or, we will be generating marketing activity over a significantly longer period of time to the same target group of customers."