Australian Data Managers Debate Quality Concerns

The pursuit of data quality requires solid, clearly defined structures, an understanding of what a firm is trying to do with data—including support of analytics—and, in some cases, careful reconstruction.

Maria Milosavljevic, chief information officer at the Australia Transaction Reports and Analysis Centre (AUSTRAC), speaking at the Sydney Financial Information and Technology Summit on August 17, said the need for solid data quality structure and uniform data quality depends on what an organization is trying to achieve. AUSTRAC is currently rebuilding its own data foundations.

"I don't think there's a one-size-fits-all approach. For our organization, we have a regulatory part of it and an intelligence part of it," she said. "Those two areas work together but they have different mandates and thus different data needs. Data quality affects them in different ways. They also have various types of action that they take as a result of the data, and that then determines what the impact of data quality is on the decisions that have to be made."

Data quality means different things for different industries, added Dipen Mehta, global head for data, architecture and innovation at Standard Chartered.

"For us, lots of our data is actually customer-driven. It originates from the customer," he said. "So verifying the quality of that can only go so far. For us, it's actually more of a lineage. When we talk about data quality in a bank, we are talking about ensuring and understanding the lineage of that data from the point at which we sourced it, all the way to how we are using it for applications and business processes. While the definition may be slightly different, we are big believers in data quality."

Paul Ormonde-James, chief data officer and head of business intelligence of NIB Health Funds, a health and medical insurance firm, said financial institutions should treat data as they would any other asset within the organization. "As a chief executive or even a board, you can't walk away from the fact that if you don't have good data quality, your analytics are garbage," he said.

According to Mehta, a lot of the problems around data, and in particular data quality, tend to be caused by bad architecture. This could mean that systems are not linked or a lot of manual processes are involved, for instance.

"The lack of architectural code leads to data quality issues," he said. "But now there is a real value proposition around data, and it's not about monetizing your data but understanding and using the data you have in new ways."

A year ago, Standard Chartered implemented a data quality management framework (DQMF), a set of governance principles for process and business owners. "It's the executives that own the data who are responsible for its quality," said Mehta. "DQMF is governed by the owner of the data, so if it ever goes pear-shaped that business person is in trouble."

Meanwhile, Ormonde-James said it is the people who are using the data that are ultimately accountable for its quality. "Data is used cross-functionally," he said. "You need someone who takes a lead and then strong communication to make sure other people know what is going on. Your governance is about the structural taxonomy that you have in the organization.

"Define who is the lead and which subsets have a vested interest, and make sure whoever is taking the lead is talking with the subsets. Nobody gets this right, and it doesn't matter if you are big or small. But it does work by setting those parameters, and making clear who does what, at what level and at what frequency."