*Implementing Database Marketing Requires Communication, Experts Say
Fleet Bank, the Boston financial giant, had just completed two large acquisitions when it began creating the data warehouse in 1996. The system took nine months longer than anticipated to implement, but even then -- three years down the road -- the company still ran into problems, said Randall B. Grossman, who spent several years overseeing the construction of Fleet's data warehouse and now is undertaking a similar endeavor as senior vice president and director of customer and database management at Bank One.
"In the original plan, 1998 would have been the ramp-up year, and in 1999 we would have been enjoying the success," he said. Instead, 1999 has been the ramp-up year for the program, and the timing delays were not communicated clearly throughout the business units of the bank. "What that meant was that we should have been managing expectations downward, often just as we were getting ready to celebrate a major milestone."
Fortunately, he had orders from top management to negotiate flat rates with the technology partners he was working with, and the project did not go over budget, even though it was almost a year overdue.
The bank uses the warehouse for several functions, including to calculate when to offer certain rates on certificates of deposit and to determine when to cross-sell other products.
Once the data warehouse was implemented, the company also ran into problems with the individual data marts within the warehouse. He said the company tried to drive campaign management off a database that was optimized for business analysis, so some of the elements were not ideal for marketing. As a result, Fleet is moving toward having two different databases, one for management analysis and another for marketing, he said.
"It's better to have several marts with redundant but focused data structures than to try to do too many things with one mart," he said.
There also were problems with data quality, he said, although only about 1 percent to 2 percent of errors were caused by incorrect data entry. More often, data errors were the result of faulty loading procedures in which data from the branches were sent in incomplete or using inconsistent metrics, which result in missing or corrupt data in the warehouse.
To avoid this problem at Bank One, Grossman said the company has created a data dictionary that has a one-paragraph definition for each of the thousands of bits of data that can be entered into a database. Many of these definitions include cautionary remarks about potential errors in filling in a data field.
Meanwhile MediaOne, the Englewood, CO, broadband communications company, also experienced some problems implementing a data warehouse system, according to Mark Voboril, vice president of consumer marketing and information. The company is seeking to optimize the 150 million to 200 million contacts it makes with its 5 million customers each year as its tries to cross-sell the four product lines the company offers: analog video, digital video, high-speed Internet access and local telephone services.
Although database marketing is showing significant results for the company's Western regional division, Voboril said a lack of communication with sales representatives led to some underperforming direct mail efforts. Some mailings that went out were getting a 6 percent to 7 percent response rate, he said, but the company was losing those results when prospects called because the representatives were not aware of why the prospects were targeted with certain offers.
"We did it right. The database was doing what it was supposed to do," he said. "We were hitting the right people with the right offer at the right time, but we found that we had to communicate to the reps why these people were being targeted."