Helen Duke, Centrelink's senior IT development manager for the data systems development section, said the organization is undergoing a project to rationalize the number of presentation tools it uses and to replace its current data warehouse platform.
Centrelink's data warehouse environment consists of a "number of presentation products" including Cognos Inc. and Hyperion Solutions Corp., which process reference data and business rules. The data warehouse infrastructure consists of 100GB on NCR Corp.'s Teradata used for fraud detection, 500GB on DB2 on Sun Microsystems Inc.'s Solaris, and 189TB on MVS cartridge.
Duke said a new tender is due for release and is waiting final approval.
With 1000 service delivery points, 25,312 staff, 26 call centers, 87 million letters per year, 6.4 million customers, and 440 billion data items, Centrelink's data quality issues are ongoing, but according to Duke, are not dissimilar to other enterprises.
"We're doing a lot of work on data integrity routines," Duke told a data quality conference in Sydney last week.
Don't try to hide the data quality problems; show business the implications of poor data quality, she said.
Duke believes many business problems with data quality can be avoided.
"You must have a champion with clearly defined responsibilities," she said. "Remove turf wars; don't have a data quality team, because data quality is everyone's responsibility."
Duke stressed that business, not IT, "owns" data quality, because the key issue to better exploitation of data resources is making people aware of what is available.
"It's hard to convince IT they need to let go, but giving control to business means [its execs] are less likely to ignore these issues," she said, adding that business often doesn't realize the wealth of information that is hidden in their data.
"If you can measure and quantify the cost to the business, you can raise the profile of data quality [so] you need to talk in business terms and not technology; then you are more likely to get engagement from senior management."
Duke also preached good governance of the people entering data, because "even if you get all the technology right, people could still screw it up".
"Unless you work in automated environment, people will be a factor [and] an ongoing process for monitoring data should also be in place," she said. "For example, we discovered that about 400 people who are male had Ms in their title and 450 women had Mr in their title."
Linking reference data and business rules in isolation won't help identify gaps in technical processes, because "it's not always IT processes causing problems, often it's the business processes", Duke said. "Ensure you understand the relationships between the data."