Nicola Tavendale

Service and product development: Priority lists for leading FX algo providers

February 2024 in Industry Views

Institutional traders shared their predictions for the coming year in one of the largest cross-asset surveys of its kind, the 2024 JP Morgan e-Trading Edit. Key sentiments which will be of particular interest to FX algo providers and users alike include the expectation that volatile markets will be the greatest daily trading challenge, followed closely by liquidity availability and workflow efficiency. In turn, traders expected electronic trading in FX to further increase this year, then rising to a predicted 73% for 2025. These wider industry trends all appear to be working in favour of the use of FX algos, but to what extent are algo providers gearing up to ensure clients have the right tools and support to confidently opt for algo execution in this evolving trading environment? Nicola Tavendale writes.

According to the JP Morgan survey findings, the significance of liquidity availability is creeping back up to regain its number one spot, having risen in ranked importance from 22% to 24%. It was only overtaken as a leading concern among traders by volatile markets at 28%, while workflow efficiency also rose to 13% from the previous year. According to Scott Wacker, Global Head of FICC e-Sales at JP Morgan, 100% of survey respondents predict to increase their electronic trading activity over the coming years. “It’s an exciting time for the electronic and automation space right now, as we look to offer clients added choice of execution options,” he added.

For FX algos however, the previous boom in demand seen during the Covid period has subsided and volumes now remain at a steady rate, says Vivek Sarohia, Global Head of Alternative Execution Services at HSBC. In addition, he notes that in terms of client type, HSBC is actually seeing less usage by the traditional real money client base and more from the systematic hedge fund, corporate and banking sectors. “Transparency of an algo’s performance, both in-flight and post-trade, are important to all client types, so we continue to invest in algo analytics and post-trade TCA,” Sarohia adds. “We are also looking at how the algo suite can provide greater spread capture, reduce market impact and increase speed of execution on open-ended orders. An effective way to achieve this is through greater internalisation and, in response, we have been rolling out our new internalisation methodology to clients. We refer to this new methodology as ‘floating IX’, which allows the client to act as a liquidity provider to HSBC, who is the principal, all in a fully anonymous fashion.” 

According to Sarohia, this new functionality also allows client orders to be placed at levels inside the spread with HSBC and to float dynamically with movements in market prices. He explains that due to the significant size of HSBC’s FX principal franchise, this enhancement increases the likelihood of a fill order, with minimal impact on mark-outs. HSBC’s other key focus in 2024 is the deployment of its FX Basket Algorithm, says Sarohia. “The basket algo has been designed off the back of client demand where they wish to execute a portfolio of FX orders, typically across correlated trading pairs, but in a way that reduces the execution risk of the orders impacting each other, as well as reducing overall transaction costs,” he adds. “The HSBC basket algo allows a client, such as an asset manager with different investment portfolios, to place a portfolio of FX orders, and then nets and splits the constituents into directly tradable pairs before calculating optimal execution trajectories. It does this by minimising constant absolute risk aversion (CARA) utility, taking into account portfolio covariance matrix, non-linear market impact and varying market volume.” 

As a result, HSBC’s basket algo provides three key benefits for clients, says Sarohia. Firstly, it ensures the netting of an FX portfolio to avoid unnecessary trading activity, thus reducing transaction costs. Secondly, he notes that the algo’s awareness of correlations between the underlying pairs helps it to execute at an optimal pace which reduces market impact on the basket, while the third benefit for clients is that the algo also minimises the weighted average slippage from the decision prices while controlling the market volatility risk.


Meanwhile James McGuigan, FX Algo Product Manager at Citi, agrees that high levels of internalisation is a popular topic for algo providers to focus on – and for good reason. “Internalisation, along with intelligent usage of external venues, can be key to achieving a better quality execution with lower levels of market impact and we continue to improve our performance in this area,” he says. “However, our clients have been telling us that they are not yet convinced that achieving a high proportion of internalisation on their orders is necessarily always beneficial. Using both our own in-house analysis and also working in conjunction with third-party TCA providers, we are aiming to help clients better understand the quality of internalisation they achieve so that they can add this important additional dimension to the way they assess each provider’s execution.”

In addition, improving execution analytics continues to be a focus for clients, with pre-, live and post-trade being more or less important, depending on each client’s execution style, McGuigan explains. “Clients who actively monitor the execution of their algo orders appreciate the market colour that we provide and are increasingly looking for more sophisticated data from the algo itself, in real time, to explain why it is executing the way it is, including when it decides not to trade,” he adds. “This information can help clients determine if they need to make in-flight changes to meet their execution objectives.”

Data-driven decision making

Alexis Laming, FX Algo Trader at Crédit Agricole CIB (CACIB), adds that as CACIB has a client-driven approach, their focus tends to be that of that bank’s as well. The exception, however, might be the limitations in the use of AI when it comes to algos, as Laming believes it is important to be fully able to explain to the users how the algo will behave and why CACIB believe it is the best strategy. “Otherwise, if our clients need new products or tailor-made tools, we are happy to discuss it. For instance, in order to meet our client needs we are currently working on deploying NDF algos as we have a good footprint in this space,” he says. 

According to Laming, clients now have a lot more experience in using algos and are now fully aware of the main pros and cons of using different strategies, but still tend to lack the data needed to make an informed decision pre-trade. He adds: “Algo providers should help them have all the cards in hand to choose the best strategy, based on the execution needs and market conditions. Data is key to assessing performance and to using the best algo given the market, but this is often expensive or hard to digest for some buyside users. We believe that talks between users and providers around the use of third-party analytics is helping the algo ecosystem to grow with increased transparency.”

Analytics and data, and how clients use that information to execute their FX orders, is also at the forefront of HSBC’s FX digitalisation strategy, says Sarohia. “Understandably, clients do not want data which is not relevant to be pushed to them, but rather to be consumers and pull what they need from banks, when they need it, in formats which they can consume efficiently,” he adds. “HSBC has addressed this change through the development of our next-generation trading terminal, AI Markets. AI Markets uses purpose-built natural language processing (NLP) to allow clients access to data which was previously extremely costly to acquire, fragmented across multiple sources, or just unobtainable.”

As a result, he adds that clients can now gain access through AI Markets to HSBC’s real-time and historic cross-asset data sets, generate bespoke financial market analytics for themselves and browse the latest market insights, depending on their exact needs and preferences, through a choice of either web-based platforms, commercial chat channels or direct API. “HSBC partners with clients to address their data needs and provide a dynamic service able to respond quickly to changing requirements of clients and the marketplace,” Sarohia explains. “We see this adaptability as a key element in providing confidence to our client base to partner with HSBC on their execution.”

Looking ahead, HSBC is also fully supportive of the aims of the FX Global Code and Sarohia expects this year’s FX Global Code review will continue the drive for greater transparency and uniformity, especially around disclosures and TCA. Laming adds that as algos are gaining market share, the review will certainly impact the algo market and predicts a similar scenario to the BIS report on execution algorithms from 2020, when an algo taxonomy was proposed to help algo users navigate the different strategies from all their providers. “We use it to explain what is to be expected from each strategy we have in our suite,” he says. “For us, the path towards more transparency is always a good approach.”

Future evolution

In terms of the wider evolution of the FX algo market, Sarohia notes that some believe artificial intelligence will be the natural evolution for algo development – with even more decision-making of the algos being determined in real time through machine learning, rather than pre-programmed parameters. “We shall see if this rings true, as I think the FX algo market is still some way off in this respect,” he adds. Instead, it was actually internalisation which proved to be one of the leading drivers behind clients choosing to use HSBC’s algos in 2023, Sarohia notes. “The opportunity to access the depth of liquidity that a firm like HSBC can provide, and being able to execute algorithmically, significantly reduces the market impact of a client order, and over 90 percent of orders sent to us requested access to HSBC’s internal liquidity pool,” he adds. “Our new methodology for internalisation has seen a marked improvement in the internalisation rates of client orders, rising from average 20pc to nearly 50pc in certain pairs, depending on the time of day, without any significant impact in mark outs.” 

Laming agrees that liquidity provision is indeed extremely important when talking to algo users. He explains that clients need to understand precisely where the child orders will be sent, and why. “Optimising liquidity is part of our day-to-day work as it leads to increased performance for our clients, which is our main goal. On the other hand, internalisation is not very well defined across market participants. Is a deal causing a price to be skewed on some venue really internalised? Perhaps the Global Code review could help define a market wise definition,” he concludes.