The secret to customer data integration (CDI) success, according to Dave Frieder, vice president of IT at XO Communications Inc., is breaking it into "bite-sized pieces."
"If you try to take over the world, you'll be sitting there for months in the analysis stage," he said.
Instead, Reston, Va.-based XO began its CDI project by matching basic customer information across disparate systems. By implementing an identity management system as the first part of a long-term CDI initiative, XO was able to improve its customer information match rate between systems by about 50% in a matter of months, Frieder said.
The driving force for the CDI initiative began a few years ago, when the national telecom provider realized that it had a serious problem correlating information in its three core databases.
XO provides its business customers with voice and data services, including local and long-distance calling services, DSL, fixed wireless, data networking, Internet, long-haul network assets and a recently launched Voice over Internet Protocol service. The company serves approximately 300,000 customers nationally.
XO runs a CRM system from San Mateo, Calif.-based Siebel Systems Inc. for customer, sales and marketing information; Plano, Texas-based MetaSolv's Telecom Business Solution for provisioning and activation; and Eden Prairie, Minn.-based ADC's SingleView platform for billing. In the past, when a new customer came on board, information was entered separately into all three systems.
"It's a classic problem in telecommunications companies that creates obvious problems," Frieder said.
That is, if a customer wanted to add or change services, say add new lines or a new location, their customer record couldn't always be easily found in all three systems.
XO launched a project to automate the flow of transactions between systems, using a commercial enterprise application integration package from Vitria, a business process integrator based in Sunnyvale, Calif. Now, new customers are entered only once and information is propagated throughout all systems.
The initial project helped, Frieder said, but it also uncovered another problem. For existing customers, who were entered into the systems before the new system was installed, disparate customer data was still difficult to correlate.
Not only was this a problem for customer service, it was a problem with an obvious cost attached. Every time an order couldn't flow through the systems, it took valuable time to manually correlate the information.
Getting buy-in from across the organization to fix this problem was easy, Frieder said. Everyone wanted a solution.
XO did a quick scan of the marketplace for an appropriate CDI product and narrowed its choices to Purisma and Informatica, both based in Redwood City, Calif. A proof of concept project by Purisma, using samples of XO's own data, closed the deal. Though Purisma was a relative newcomer to the market, with its Customer Registry v1.0 only generally available in May 2005, Frieder felt confident in the product.
The first goal was to improve the match rate of the basic customer account information across systems. XO quantified this match rate at the beginning of the project by analyzing how many positive matches the system came up with when comparing customer data across all three databases.
Next, Frieder explained, Purisma met and trained XO's IT developers, a staff of three people. Then, Purisma facilitated joint workshops with XO's IT department and internal business users to figure out the appropriate business rules and fuzzy logic for matching customer information.
Once a set of rules was drafted, they were tested on the data and the results were analyzed. This process was repeated three or four times to tune the algorithms that the Purisma system would use.
Then, XO implemented the Customer Registry system, which created a customer identity management system to cross reference customer information between the databases. None of the existing databases or interfaces changed -- they simply used the new customer identification system to match and reconcile information.
The result was a match rate that improved by about 50%, Frieder said.
"What's represented as the customer name in each of the systems didn't change," Frieder explained. "But we now know the logical association."
XO will eventually implement an automated way to change the data in each system, but it's a low priority, Frieder said, because now they have a way to match the information.
Working with customer information was phase one. Now, Frieder's team is working with the product information, and next they'll look at product features and attributes.
"You have to work your way down through that chain, to ultimately clean up all of your data in all of your systems," he said. With hundreds of data elements in XO databases, it's a process that he expects to go on for years.
XO's keys to CDI success?
The No. 1 priority in undertaking a CDI project is to develop a specific problem statement and a metric that you can use to measure the success of the project, Frieder said. Next is to know your data and have a strategy for how you're going to attack it in manageable bite-sized pieces. And finally, "execution and perseverance" are critical for any data management project, Frieder said.
"As hard as we try to write systems in a way that won't allow data to be different, there will be different data," Frieder said. "[This project] didn't make us cut people -- it allows us to do more work with the same number of people and do other kinds of work as well."