For Eric Hirschhorn, data quality is important, but in his view, the conversation about data quality focuses on the wrong things.

“We woke up one day and someone sent me a feed of customer records, and the data quality was bad,” he said. “What does that mean? Was their zip code bad?”

Hirschhorn, chief data officer at BNY Mellon, delivered the keynote address at the North American Financial Information Summit on May 17 in Manhattan. He said the problem with data quality lies in the industry’s data processes, specifically those that allow for things like improper zip codes to be entered the first place.

“We walk around talking about these data quality problems, and we put a lot of technology around them. What we really start to lean into is the fact that we probably had an operating model that needed to change,” he said. “It’s like keeping food fresh in the refrigerator—it takes work, and it takes an approach.”

Hirschhorn said the focus on bad data quality should shift to bad processes. “At times we confuse data quality with insufficient controls at capture. As an industry, we sometime think we have a quality problem, but often we really have a process problem,” he said.

The solution, he said, is data centricity, where applications are built around a central data architecture.

“I didn’t invent the term data centricity, but applied to financial markets, it’s transformational,” he says in a separate interview. Other industries and markets that have operated from a data-centric perspective from day one contrast with the capital markets, which are continually transitioning from manual processes.

“It creates a really interesting juxtaposition because everybody says, ‘Let’s emulate the new people using new technology,’ and some of that great new stuff turns out to be the result of being … data-centric by design,” he says.

Hirschhorn says in the financial industry, data typically moves through a process-centric model. For example, as a client relationship matures, information will move from teams responsible for contract negotiations, to those performing know-your-customer checks, to onboarding teams. Technology solutions are created for each group to manage incoming data and process it. But in a data-centric model, teams and systems merge.

“If we start to think of it in a data-centric view—about building this client relationship from a data perspective—we tend to merge the systems and merge the teams, and we spend less time sending the file proverbially from group to group,” he says. “If you can concentrate on building one customer record in one system, and everybody participates in that supply chain or manufacturing process, it gives you decidedly different outcomes than having five groups collaborating like an assembly line.”

Barry Raskin, head of data practice at TickSmith and former president of Six Financial Information, says the issue of data quality isn’t always straightforward.

“Data quality is in the eye of the beholder,” he says, adding that metrics for assessing data quality could be in its completeness or its timeliness. “If I am concerned about the timeliness of a certain corporate action, some providers might provide that corporate action data when it’s initially announced but not finalized,” he says, which can lead to a provider having to issue correction messages that contradict initial data.

“Quality is a relative term, and that doesn’t mean that there are people out there doing crappy data,” Raskin says. “I think it’s also: What’s your measurement? And what’s important to you?”