Liquidity bridges are, arguably, the most important and overlooked pieces of FX brokerage infrastructure. This may be partly due to how they originally emerged as relatively simple middleware between legacy trading platforms and liquidity providers.
But as the industry has evolved to adopt different business models and execution practices, connecting to multiple LPs, and offering greater asset variety to a much more diverse client base, the demands placed on this cornerstone of brokerage infrastructure have changed.
As functions like bridging and aggregation increasingly merge with risk management, intelligent order routing, and analytics; the traditional FX technology stack is being rethought.
Modern brokerage systems must handle pricing, liquidity aggregation, matching, order routing, execution, internalization, hedging, exposure management, real-time monitoring, data management, and reporting, among other things. And the systems responsible for handling these needs must be able to communicate with each other seamlessly for brokers to be able to respond to the overwhelming complexity of modern trading practices.
Due to a combination of evolving trader mores, and increased vendor competition, the industry is being forced to reconsider the modern FX brokerage technology stack and to decide how to piece together its various layers in a manner that guarantees competitiveness and avoids future bottlenecks.
Increased competition, lower prices
Parallel to this, there’s also been an explosion in vendors, each with their own areas of specialization. This increase in competition has led to a race to the bottom in terms of pricing, and the strategic expansion by vendors into related infrastructure that can be bundled into more attractive packages.
We’ve seen companies that were originally platform providers bringing their own bridges to market and backend vendors venturing into platform provision. This suggests that as competition compresses pricing, the “all-in-one” approach is gaining ground among technology providers.
Are liquidity bridges created equal?
This raises the question of whether bridging technology has become commoditized. Are we dealing with an interchangeable component that can be thrown into a package to sweeten the deal, or can one vendor’s technology offer a competitive advantage over another’s? As with much in life, the answer is rarely a simple binary.
Indeed, the very idea of a bridge as a simple link between brokers and their LPs is insufficient today. Beyond being a connector, today’s bridges incorporate a variety of crucial features that include risk monitoring, price analysis, and decision logic that are more like liquidity aggregation, execution, and risk management layers, rather than simple bridges.
One of the ways that vendors are seeking to differentiate themselves is by packaging liquidity aggregation and bridging technology with risk management software, as this pairing covers many of the features the dealers using it require.
Essentially, we’re not just comparing bridges, but rather a collection of systems designed to monitor exposure, segment and route orders (A/B/C-book), intelligently distribute flow (by price, depth, past performance, relationships etc.), with low latency, while being easy to customize, and also providing team members with useful analytics and control. So, no, not all bridges are created equal, but today this is more to do with the assortment of other systems they’re necessarily connected to.
The very idea of a bridge as a simple link between brokers and their LPs is insufficient today
All-in-one or best-of-breed?
In the past, the question was whether to outsource technology needs to third-parties, or to develop in-house. This issue has been put to bed in recent years as it’s become increasingly clear that the overheads and technical debt accrued by in-house solutions run contrary to the core competencies of most brokerages. Today, the burning question is whether to stay within a single vendor ecosystem, or to mix and match “best of breed” solutions.
The above question is complicated further by the fact that, unless you’re dealing with a new entrant to the space, brokers have their own legacy systems and existing relationships, so these businesses are often forced to add new pieces of technology to their stack on a case-by-case basis, which is why apples to apples comparisons are often difficult to make.
Even for an idealized brokerage that’s a blank slate, there are trade-offs between sourcing key infrastructure from a single vendor or piecing together different best-in-class components. These trade-offs must be weighed by decision-makers who have a long-term view of the business’s needs and goals.
For example, all-in-one solutions tend to have lower latency than disparate components connected via APIs, while being both simpler to manage and quicker to deploy. On the other hand, vendor lock-in, and lack of specialization in certain components may be a concern for brokers.
The pick and mix approach means that a brokerage can be constructed in a highly customized way, with each component being specifically fit for purpose. It also offers the business in question the flexibility of multiple relationships that can be leveraged to get the best deal. However, despite this, disparate components increase complexity and management overheads and are generally more costly.
Where’s the brain?
For modern brokers the technological brain of their respective infrastructures is a collection of back-end systems that must work in concert in order for the venues in question to be able to manage both internal and external risk, and to facilitate the execution of trades in a seamless way at the current best price available to the end client.
The ratio between front-and back-end tech has, over time, moved in favor of a very thin slice on the front-end, essentially just a trading UI, with back-end systems swelling to incorporate many more intelligent features.
In this article, we’ve focused more on the union of risk management, aggregation and order routing technology that instantly validates whether a client is within their limits, decides on how to execute the order, finds the best LP to forward it to, and then updates the system, all in fractions of a second.
But this brain isn’t just limited to execution and risk. A host of other client segmentation and behavioral analysis tools have also gained ground. These allow brokers to obtain a granular view of their client base, tailor communications to different segments, and interact with them in real-time depending on what activities they’re engaged in on the platform.
While latency isn’t as much of an issue for these features, it does greatly increase the complexity and interdependency of back-end systems. For the price of this complexity, brokerage teams are now able to utilize the massive amounts of data their businesses generate in a manner that simply wasn’t possible even a few years ago.
In some ways, this means the modern brokerage hosts something like a distributed brain, one area concerned with liquidity aggregation, risk management and execution (the “bridge” being integral to this), and another concerned with client intelligence (feeding into marketing, sales and retention initiatives), as well as other “lobes” involved in back-office accounting activities, compliance, and reporting.
APIs connect the lobes
With this in mind, it becomes clear that most modern brokerages will necessarily have to pursue a hybrid approach in which certain best of breed solutions that are crucial to the overall brain as described above will have to be selected and integrated with others.
CRMs are a case in point. This is a specialist component, as critical to the smooth-running of a modern brokerage as is the blend of liquidity and execution technologies outlined previously. And the vendors with expertise in the former are rarely the same as the ones with expertise in the latter.
For this reason, APIs, and specifically APIs that are left open and comprehensively documented by their respective vendors, are increasingly important as they allow systems to be integrated in ways that serve each individual business, and also allow firms to perform their own custom integrations.
This hybrid approach has gained ground in recent years, with the combination of unified platforms, risk management, aggregation and execution technologies offering the speed and reliability that this crucial layer demands, while other mission critical parts of the brain can be sourced and integrated according to the organizational needs of each business.
Boxes to tick
So, considering all this, we can discuss a few boxes to tick when sourcing these various technologies. Open APIs are one of these items on the account management and client intelligence side of things.
Another item is the ability of brokers to customize, modify, and retain control over how they segment client groups, devise and evolve their order routing rules, and apply differential pricing to different client groups, as well as the ability to implement these changes in business logic rather than in code.
AI probably deserves a section in its own right as the absence of AI tools isn’t a deal breaker in and of itself, at this point, but this is a rapidly developing situation that’s changing from one quarter to the next.
In terms of handing over business-critical processes to AI, the industry is still divided on this. The growing demand for hands-off automation contrasts with certain understandable reservations due to the massive risks involved in handing over control of routing, execution, and risk management to artificially intelligent systems, rather than having them serve as dashboard telemetry for human teams.
Our own approach has been to focus more on what AI can do to better understand client needs, providing human teams with a more surgical understanding and reach at scale, as this holds great potential to unlock efficiencies and new approaches to client outreach that can level the playing field, making smaller outfits much more competitive.
In the long run, AI will probably be the technology that necessitates yet another rethinking of the brokerage technology stack due to how helpful it can potentially be in a wide variety of both client-facing and back-end use cases. At the moment, this is a moving target. A great deal of experimentation and real-life testing is currently taking place, without any objective best practices having been settled on as yet.
