There’s a knowledge gap at financial firms. A firm may have all the data in the world, but its employees often don’t know they have it. And even if they know about it, getting access to it may be hampered by lengthy approval processes that delay access to data. In many cases, this data already exists within their organizations, and is already being paid for, but users seek fresh datasets because they don’t know the data already exists.

In short, firms don’t know what they know.

This challenge doesn’t have a single cause—or solution—but rather, is a multi-faceted problem, comprising commercial issues, organizational structures, and technical obstacles.

“With current data licenses, there is no central way for people to say, ‘What data could solve which problems? Are people aware of the data? And if they were, would they use it?” says Suvrat Bansal, former chief data officer at UBS Asset Management, who has founded a startup data company, Stellar Data Labs, aimed at making data more easily discoverable within financial firms. “These things are missing in organizations today. But once we do that, let the floodgates open and let the knowledge begin.”

We’re so busy focusing on things like data quality that the data is not reaching the people who need it. … We spend so much time hiring smart people, but then we handicap them by not giving them the data.

Suvrat Bansal, Stellar Data Labs

At the core of the problem is that firms are not structured in a way that makes data bought or created and stored in one business line easily visible—or accessible—to potential users in other areas, even though it could have beneficial uses in many parts of an organization.

“If you talk to any organization today, they’ll say they’re investing a lot in data. But if you ask the same person if it’s easier to get to that data, they pause. We’re so busy focusing on things like data quality that the data is not reaching the people who need it,” Bansal says. “There’s a growing distance between data and users. It can take between three weeks and three months for data access to be approved by multiple IT teams. We spend so much time hiring smart people, but then we handicap them by not giving them the data.”

The Netflix model

Instead of organizing data by business functions like risk or finance, Bansal’s platform organizes data by themes—such as client centricity or sustainability—that cross different business areas and functions, and recognizes that people in different roles performing different tasks need their data made available in different formats. Some need it in spreadsheets, others need it in a format that they can integrate it into artificial intelligence- and machine learning-based applications.

“These are concepts that cross the boundaries of functional data. If we don’t give access to that, how do you service clients correctly or perform ESG-related tasks correctly if you can’t give teams that view?” he says. “By structuring the product by domains—such as client centricity or finance—we’re giving users the ability to access data within the context they want to access it in.”

Bansal says he took his inspiration from retail marketplaces, such as streaming video service Netflix, where users can search for movies they’re aware of, or browse movies sorted and categorized by Netflix, and watch the movie on the same platform. In the same way, data consumers could search for data they don’t know exists within their organization, and firms could control who can access data by applying permissioning in the same way Netflix viewers set up parental controls to block certain movies and shows.

Once the platform maps a firm’s datasets, it presents the data to users in the format they want to access it, such as in tables or charts for analytics, providing a more suitable tool for viewing data than spreadsheets, which, according to Bansal’s research, remain the most widely used way of viewing data outside of data workstations.

“I estimate that 90% of people still analyze data in spreadsheets, while 5% use charts, Tableau, etcetera, and 3% use it in self-service applications, and an emerging 2% is used by AI and ML teams using notebooks,” Bansal says. “But a dataset may have 150 million records. You can’t do that in a spreadsheet.”

‘Current technologies just aren’t working’

Others familiar with this space say Bansal’s efforts are needed to help firms understand their data, but warn that it could prove a complicated and costly endeavor.

“Data discovery is a big area, and current cataloging technologies just aren’t working,” says Brad Schneider, CEO of Nomad Data, which provides a platform for matching end users’ data requirements against datasets offered by alternative data providers. “We are continually getting questions from CDOs, such as, ‘What assets do we have, and what are they useful for? How do we make our own data discoverable? And how can we hope to do that with external data if we can’t do it internally?’ Clearly that’s a pain point for many types of organizations.”

Schneider notes that being aware of what data a company already has—and is paying for—isn’t just good governance; it literally pays off to use a dataset you’ve already paid for to perform your research or a valuation—that is, if you know the data exists and can easily find and access it—rather than paying for another set of duplicative data.

Making data discoverable and usable faces organizational challenges as well as cost challenges. Most organizations still don’t have a CDO, and still need to get management buy-in to empower an individual to oversee this before they can even hope to begin the tasks of making data discoverable and accessible, Schneider says.

“A centralized data catalog is a dream, but it only gets you part of the way there. The hard part is, how do you plug into all the data types,” which may be stored differently, have different metadata, or where a firm may experience internal turf wars over who can access what datasets. “It’s really impossible to take a one-size-fits-all approach, because you keep finding datasets that make you re-think how you do everything, and you need to build a very competent team to handle different areas. It would be a very expensive proposition to do that.”

Thankfully for Bansal, though he decided to bootstrap the company with his own money rather than diluting his stake by seeking venture capital investment, he has already signed up three banks—one in Europe, one in the Nordic region, and one of the top five US investment banks—to put his platform to the test.

He says the offering has broader appeal to other industries, and that he hadn’t intended to focus initially on large banks, but that its sweet spot is organizations with complex data structures—and that these problems are at their largest and most complex in large financial firms.