​Bigger is better, right? From modern financial markets to the regulations that govern them, and not least the financial services firms themselves, the technology that keeps them running, and the datasets they consume to power every function from price discovery, analysis, risk management, and settlement, everything is significantly larger than just a decade ago.

But size also brings with it greater complexity. Understanding a firm’s overall exposure to market, regulatory, and IT risk each present enormous challenges. But new tools offer an opportunity to unravel some of that complexity, and reduce inefficiencies that have arisen as a result.

Knowledge graphs visualize relationships between individual data points behind the scenes like a matrix or three-dimensional grid rather than displaying information in a traditional linear or relational manner. They are already used by data providers to spot associations between datasets, creating a more comprehensive view of an entity or security, and adding value to their offering. Now, though, financial firms are starting to deploy tools that use the same underlying technology to solve some of these specific business challenges, ranging from IT deployment activities across complex organizations and structures to systemic market issues.

One area that the technology is being used is to assess the impact that a potentially minor change to one system has on all the other systems with which it interacts. These tools can be used to identify and mitigate any unforeseen consequences, without having to first conduct a fishing expedition to identify all the systems that would be impacted, and then undertake extensive testing to assess the severity of any impact.

“The problem has existed for a long time. The challenge is, if I make a change here, what’s the impact of that change elsewhere,” says John Bottega, president of data management industry association the EDM Council. “We, the financial industry, have looked at impact analysis linearly in the past. But now, the impact is multi-dimensional, so by looking at these relationships in a non-linear, 3D manner with knowledge graphs, you can get a better view of data across multiple applications.”

For example, Philip Dutton, co-CEO and co-founder of UK-based data lineage platform vendor Solidatus, says that a developer at an end-user “added a couple of decimal places of precision” to a rate field in an FX trading system. The changes were tested, but the system ultimately crashed because a downstream legacy system had fixed-length fields and could not handle the improved “precision” that the developer wrote into the system. As a result, the firm had to manually make “hundreds of thousands” adjustments.

“If that happens on a day when there’s a big move in the [British pound/US dollar] cable rate, that could translate to millions in losses,” Dutton says.

While the system would not prescribe how a firm should respond, it highlights deviations from a firm’s prescribed ruleset, especially where problems are—or are exacerbated by—manual bottlenecks. The aim of his company’s platform is to flag an error like this pre-development to stop the error from weeding its way into the live production environment. “A change in the number of characters shouldn’t stop one of the largest banks in the world from trading FX. But these things happen all too often,” Dutton says. “Banks have trading days that just halt because of minor system errors.”

Beyond technical impact assessment, the same tools can also aid compliance efforts—and potentially help unravel complex interdependencies—by mapping the relationships between every dataset.

For example, while not able to prevent the 2007 financial crisis, these tools would have made it easier for market participants to spot risks, evaluate exposures, and respond to them faster, such as the “multiple tentacles” of risk at Lehman Brothers that took years to unwind, says Bottega, who was then chief data officer at Citigroup, and—as Wall Street unraveled the fallout from the financial crisis—subsequently became CDO of the markets division of the Federal Reserve Bank of New York.

“What knowledge graphs provide is a view into a network of related and dependent facts and events. Given the complexity of the financial crisis, this type of analysis could have exposed various areas of stress in the system,” Bottega says. In fact, the EDM Council conducted its own proof-of-concept initiative to test the effectiveness of knowledge graphs for mapping and unraveling complex structures. The EDM Council looked at a series of derivative trades, modeled using the FIBO (Financial Industry Business Ontology) meta-model ontology and a graph database. “What this POC demonstrated was a way to quickly look at trade relationships, concentrations, over-exposures, etcetera, that are critical to someone performing a risk analysis.”

Though the concept is not new, the finance industry is realizing not just the benefits, but also “the necessity of going in this direction,” Bottega adds.

Thankfully, most of the challenges facing firms are more mundane, but still have the potential to be costly drains on resources.

For instance, the plethora of recent regulatory reporting requirements have resulted in individual business lines within financial firms implementing their own reporting solutions just to serve their specific compliance needs. Regulations that impact multiple business lines within a firm range from GDPR and privacy rules to BCBS239 governing risk aggregation and reporting, and regional regulations such as the Dodd–Frank Act and Mifid II. However, the scope of regulations means other departments are doing the same thing, unaware that a colleague in a different department is duplicating their effort—and cost—says Solidatus co-CEO and co-founder Philip Miller.

“A lot of regulatory requirements are about trying to leverage the same data for a different purpose, and so different teams end up doing similar work, duplicating efforts to make the same solutions,” Miller says, “and they are spending millions of dollars to achieve that.”

Because of the intense rollout of new regulations in recent years, some firms have taken a concerted approach in their response to new requirements. But for many, the sheer complexity of their firms and the tasks being performed is impossible to keep track of.

“Typically, the answer to why companies don’t address these issues in a joined-up way is an organizational challenge,” adds Dutton. “These large, multinational companies have grown so complex in their operational, functional, and jurisdictional alignment, that even the determination of obligation to regulation is extremely difficult.”

Divisions may have budget autonomy and a list of priorities, but they might not have an enterprise-level awareness of how the same regulations impact other areas. So they buy and build technology “in isolation and therefore duplication,” he says.

In addition, building these separately makes it harder for different areas to re-use functionality and data, leading to an “inevitable loss of consistency and quality,” Miller says, blaming current approaches to project structure and funding. “Project funding does not encourage the reuse of assets. One asset will be modeled many times and there will be no cross-pollination of intellectual property inventories, which will be in project-aligned SharePoints, provisioned to limited, project-aligned users. These documents will be wholly focused on that project and not lend themselves to reuse.”

And though responsibility for compliance or for IT-related failures may rest with one executive—whether a chief risk officer, chief compliance officer, chief technology officer, or chief data officer—Dutton says it isn’t practical to expect them to achieve full oversight across their enterprise.

“A single person can’t have visibility into all of this—they need tooling to help them,” he says.

Mike Atkin, a former managing director of the EDM Council, who is now director of the Enterprise Knowledge Graph (EKG) Foundation, says knowledge graphs solve these problems. They resolve identities to unique and permanent identifiers, express meaning using web standards, capture business rules as machine-executable models, while validating quality by mathematical axiom, facilitating reuse of concepts rather than physical elements, and—most importantly, Atkin says—recreate context.

“The truth is that everything can be expressed by combining identity, meaning, time, and source, and knowledge graphs give you that capability. What firms get is a connected inventory of everything, contextual search, flexibility in query, and access control,” Atkin adds.

A data lineage application like Solidatus not only demonstrates to regulators that a firm is in control of its data and reporting processes, and is taking their requirements seriously, but also allows it to address non-reporting related issues such as identifying similar projects so firms can better understand enterprise-wide duplication, and potentially consolidate efforts to better address technical risks.

“We use a unique versioned graph on the back-end that gives you a temporal and bi-temporal view of connected data, presenting multiple timelines so you evidence the past, impact-assess the present, and simulate the future,” Dutton says.

Once a data lineage application like Solidatus has visibility across all applications, it can understand how they interact—and where a change to one system can impact another—faster and more precisely.

To exploit this opportunity, Solidatus has just raised £14 million ($19.3 million) in Series A funding, led by private equity firm AlbionVC, and it includes two of its enterprise banking clients, HSBC and Citi, primarily to accelerate performance enhancements and develop new tools.

In addition, the funding will allow Solidatus to continue growing its headcount—the vendor expects to have between 160 and 180 employees by year-end, roughly tripling its current headcount—and to set up a US presence headquartered in Houston, with satellite offices in New York and other regions.​