Northern Trust building internal cloud data ‘marketplace’

Custodian bank Northern Trust is building a cloud-based data marketplace for users within the bank that will increase its analytics capabilities while implementing more stringent controls on usage, and may over time make the marketplace available as a data source to external clients.

The marketplace is currently being used by the firm’s asset servicing division, and is now being rolled out to its asset management and wealth management divisions, initially for analytics, risk, and reporting.

“We have an MVP [minimum viable product] of the marketplace live with a number of users. We haven’t expanded it more broadly because we don’t have a sufficient volume of data yet,” says Kelley Conway, EVP and head of corporate and digital strategy at Northern Trust.

So far, the firm has populated the marketplace with some of its internal transactional data, some of its reference data, and some external data via Snowflake, which is one of the cloud partners in the project. Northern Trust also uses Microsoft’s Azure cloud, along with elements of Collibra’s data intelligence platform, and built the marketplace front end in partnership with data mesh specialists Esynergy, though the bank has now taken over full development and maintenance of that.

We have central access and auditability, so it improves our ability to manage risk. So, you’re not only freeing and democratizing the data; you’re also controlling it better

Kelley Conway, Northern Trust

Duncan Cooper, CDO for asset servicing at Northern Trust, views these basic datasets, such as the firm’s security master and custody accounting data, as Lego bricks. And like Lego bricks, simple datasets can be connected to build more complex structures of insightful and consumable data. But that requires the right “bricks” being available, and having a way to make them accessible to users. He likens the construction of the marketplace to Amazon circa 1997.

“They had a lot of warehouses, they had a stock and inventory system that understood where everything was, and they had a portal to sell to customers. Those were the three core components,” Cooper says. “For us, the warehouses are the data mesh, which stores accounting data, custody data, and transactional data. The stock-taking capability—understanding where the data is—that’s Collibra. And the marketplace is what glues those tools together and offers them to customers.”

The key objective was to make useful data more accessible and used more widely across the bank—beyond just the data “geeks”—and, by monitoring usage patterns, be able to offer better products to help internal customers do their jobs. For example, if someone in the firm wants to analyze how clients are interacting with Northern Trust, or if someone in an advanced analytics group wants to build a model looking at customer retention, they can find and access data to do that more easily, and with built-in controls. The firm also documented case studies so staff could understand how others are using the analytics.

The project began three years ago when Conway joined Northern Trust. The impetus for the marketplace came from seeing other firms adopting data meshes while she was a consultant.

“While I was at Accenture, I saw companies moving onto a data mesh construct, and I was intrigued by it because it pushed the ownership of data away from the technology organization and into the business and operations. To me, that’s always been an issue, because the business side asks tech for data, then says the data is wrong—well, that’s not the tech organization’s job,” she says.

The firm quickly realized that establishing ownership and governance would be foundational to its main objective of making data democratized and more accessible throughout the organization.

When Conway arrived at the bank in 2021, there was no ownership of data by the business side, she says. Now, the data mesh assigns ownership but also ensures greater control of data, generating “contracts” that govern how and by whom data can be used—and also how frequently it’s used—while analysis of the data is performed in analytics “sandboxes” that are part of the mesh itself, so data is never downloaded and proliferated elsewhere in (or outside) the organization.

“We have central access and auditability, so it improves our ability to manage risk,” she says. “So, you’re not only freeing and democratizing the data; you’re also controlling it better.”

Once the firm has onboarded more datasets onto the mesh—an effort in itself to expose data into the mesh from a multitude of separate, legacy technology platforms—and expanded its internal user base, it will need to decide whether to make data from the marketplace available to external clients, and whether to also offer access to other third-party datasets.

“When we originally thought about the marketplace, we envisaged it as internal, and now we are thinking about using it externally as well, but we’re not there yet,” Conway says.

Cooper says the firm would need to spend more time on security and control features before making it available outside its walls, but can see how the marketplace could prove useful to the asset servicing division’s client base among asset owners and servicers. For example, back-office professionals may want to understand settlement details, while front-office staff at an asset manager may want to incorporate and experiment with more datasets using tools like Jupyter Notebooks to better understand market trends.

“If you look at Amazon, one thing they did was to expose their platform to resellers. So, do we maybe expose resellers to our customers [via the marketplace] to give them access to more analytical tools?” Cooper says.