Bank of America reduces, reuses, and recycles tech for markets division

In sustainability, reuse is the second tier in the hierarchy of waste management. The concept is simple: Why buy something brand new when you can repurpose something you already have for free, or at least for far less? Need a new bookcase? Sure, your local furniture store has you covered, but if you want something nice, it will cost you. You could purchase the supplies to build one yourself, though that will require time and effort, and perhaps you’re not a master carpenter. On the other hand, maybe your neighbor has a perfectly good bookcase collecting dust in their attic that is destined for the trash dump.

The question then becomes: Do you spend $300 on a brand new bookcase? Build one yourself? Or do you simply take your neighbor’s bookcase and maybe slap on some primer and paint to touch it up? Functionally, all three do the same thing: hold books.

The concept of reuse in the environmental sector involves diverting perfectly good materials from landfills. In the realm of technology, it focuses more on cost savings, resource maximization, and improving efficiency with existing assets within your four walls.

As budgets started to shrink and the regulatory requirements started to balloon, there was no other mechanism to solve this problem

Ashok Krishnan, Bank of America

Ashok Krishnan views reuse as a key component of developing a healthy tech ecosystem at a large bank. His job at Bank of America is to assess the global markets IT budget and determine what pieces should be built internally, purchased from a vendor, or borrowed—“reused”—from one of the bank’s other trading silos.

Since 2008, bank IT departments have had to do more with less. Increased regulatory reporting requirements, rapid technological evolution, constrained budgets, and limited access to talent have compelled trading firms to get creative. Constructing entirely new, isolated monoliths is no longer fashionable. When working in silos, purchasing new, third-party solutions elevates costs and reduces interoperability.

“Reuse evolved out of the consistent visibility of the enormous amount of waste that goes on,” Krishnan, head of Global Markets Platform and Electronic Trading, tells WatersTechnology. “As budgets started to shrink and the regulatory requirements started to balloon, there was no other mechanism to solve this problem.”

Thrifting

For an example of reuse in motion, Krishnan points to the bank’s repo business. Recently, the repo desk approached Krishnan’s team to develop a front-end trading system. Repo has been notoriously slow to electronify, and it has a vastly different market structure and set of complexities than equities.

But BofA had already developed a front-end trading system for portfolio and cash trading that served as an order management system and included some execution management functionality. It could aggregate, schedule orders, and send messages—functions that are not specific to equities.

Krishnan told his colleagues that equities had a platform that could deliver 90% of what the repo business required, while the tech team could develop the remaining 10% that was specific to repo.

“That trading platform has gone from equities to futures to repo to mortgages—all the businesses now have the same front end,” he says. “In the absence of this structure we have, they all would’ve built their own things.”

Then there is Bank of America’s rates team, which trades heavily in Treasuries every day in addition to swaps and other instruments. As a result, there are opportunities in rates for the bank to match its internal flow against client flow, and internalize that the same way it does in equities.

 

The equities division has an internalization engine for its dark pool, Global Crossing Exchange, and its matching engines. Rates wanted to build an internalization engine specific to its own needs. But like the trading front-end, GCX’s internalization capabilities could provide rates with most of what it needed. And building from scratch would cost more and take longer to roll out.

BofA has since deployed the GCX offering in rates, and “our internalization in rates has now quadrupled from what it was before.” What would have taken 12 months to develop internally was reduced to six months by reusing the equities offering, Krishnan says.

As a result of the successful rates implementation, the bank has recently started to lay the groundwork for another GCX crossover, this time in foreign exchange.

“We have a single GCX tech team, and they adjust it for each use case. The benefits are huge because we save on brokerage, commission, and exchange fees, and we save on spread that we pay out to the Street in terms of bid/offer,” he says. “We internalize all of this stuff into one single technology stack.”

Broadly, tech recycling in the capital markets involves taking something from equities, the most advanced asset class in terms of electronification, and applying it to other, less automated asset classes. However, sometimes the opposite route—taking a novel process from a younger asset class and applying it to an established use case—is a winning strategy.

Krishnan points to credit exchange-traded funds as an example. “Credit built a very good tool to disaggregate an ETF into its components. Equities has always been pricing ETFs as a ticker, but they never disaggregated it. We took the credit example and gave it to equities, and it’s been a huge success for equities.”

The core of the reuse mantra lies in creating a horizontal structure that enables the bank to maintain a consolidated view of technology across the organization. Silos emerge because silo owners prioritize what’s best for their own area without much regard for the other silos.

In many ways, that makes sense—everyone wants the best solution. The problem is that people are trained to believe that whatever is new and specifically built for a particular problem is the best solution. But creative thinking can go a long way.

Krishnan says that they needed to first establish that horizontal structure; otherwise, people will revert to their silos. They then created an inventory of all the different platforms and tools used across the markets division. If someone wants to build or buy something new, they have to approach his team, which looks through that directory to see if the bank already has the requested capability. If they don’t have it, that’s when they weigh whether to buy it or build it.

“If you come in and ask me, ‘What are our capabilities for RFQs?’ I can tell you exactly what we have across the whole platform. Whereas, in the past, you could go to the rates person, and they will tell you what they have in rates, but they won’t know what we have in equities,” he says. “That’s the big change.” 

Engine fuel 

One of the largest philosophical changes to occur in the capital markets is the understanding that tech and data functions should not be kept separate. A year ago, Krishnan’s team brought in the quant and data functions to secure the horizontal structure they had put in place.

Just as Krishnan’s team made the platforms and tools available across the bank visible, they wanted to do the same with data, multiple copies of which had been replicated across various silos. “It’s a very inefficient use of resources,” he says. 

A singular view of data across the silos, overlaid with a quant function, provides deeper analytics for the risk and finance teams, internal traders, the credit desk, the equities desk, and so on, all of whom can now see the goings-on in one another’s businesses. 

“The other driver for combining the quant and data functions was to ensure that data quality is something we consider first and foremost as we develop technology,” Krishnan adds. “When you are thinking about whether to build, borrow, or buy, you have to have data quality ingrained into that decision-making process.”

We started to consolidate market data into one place and ... as a result, we cut out a bunch of duplication in market data

Ashok Krishnan, Bank of America

He says that he thinks people “get carried away very quickly” when they hear about a new, flashy front-end system. The reality is that if the data is of poor quality and stuck in silos, the tech design might look pretty, but the system’s results will have limited usefulness.

In the past, it was common across the industry to build things only to later discover that the data was flawed. Teams would spend months—even years—trying to backtrack and figure out where they went wrong. Krishnan’s mentality today is to start with data quality and “work your way to the front.” Even regulators are backing this push for better data quality—something that a couple years ago they didn’t even mention, Krishnan says.

So it was that two years ago, BofA began centralizing its tech stack to consume data and remain agnostic to vendors, allowing the various teams to use the best vendor for each use case with minimal overlap and waste.

“We started to consolidate market data into one place and say, ‘If you need market data, come here, [and] we’ll sign one contract with a vendor.’ As a result, we cut out a bunch of duplication in market data,” he says.

This structure also helps when cutting-edge tools come to market, such as the recent boom in advanced artificial intelligence. As firms of all stripes and sizes have learned over the past few years, AI is only as effective as the data underpinning the models. 

“Because of the structure that we put in place where quant, data, and technology prioritization all reside with us, all of the AI implementations are also with us. This way, we can deploy it once, and everybody can use those tools.”

Reduce, reuse, return on investment

Sometimes, ROI can be difficult to quantify, especially concerning technology reuse. Nonetheless, firms need to modernize their systems and break down silos—all while doing more with less—if they want to produce high-quality data and analytics.

Krishnan acknowledges that he can’t always quantify all of the benefits of reuse. But if the bank can attract more users to a platform like GCX, it means more users are engaging with a platform that is maintained by the same team, ensuring consistent cyber protection and data uniformity in the back end. Alternatively, if they manage to retire a legacy system and transition everyone to one that is already in use, ancillary benefits emerge. Instead of having five different teams managing five different platforms, there is one unified team where everyone focuses on adding value.

Krishnan says that going back to the 2020 onset of Covid, the number of new capabilities requested—from entering new markets to adding functionality—skyrocketed while the resource pool shrank. The tech team could sense a breaking point rapidly approaching. Maintaining seven of everything—whether platforms or datasets—was taking its toll.

This led the group to the mantra of “ruin duplication.” In other words, remove all legacy duplication and all future duplication among its various platforms.

“Think of reuse this way: As a household, I’ve got a budget of money to spend, so I’m not going to buy two of the same thing,” Krishnan says. “If I’ve got a dining room table, I’m not going to buy another dining room table. That wouldn’t make any sense.”

Decades of aggressive advertising have taught us that new is better while used is cheap. And some things do need to be modernized—no one wants to use a system run on Cobol. But spaghetti gets made when everyone starts throwing in their own ingredients to suit their own palates. 

Krishnan acknowledges that in the beginning, colleagues doubted whether this horizontal structure would add value and were concerned it might instead slow the silos down. “The questions were all rightfully asked,” he says. Instead of debating them in theory, he let seeing be believing.

“Once you have a few success stories, it opens people’s minds as to what you can do with tools that already exist within the organization,” he says. “You put it all out on a shelf, and people start to think, ‘Oh, maybe we can use that.’”