More flexibility with less complexity – Streamlining the algorithmic FX trading process

To some extent this is probably the dream of every one of the hundreds of thousands of institutions that use the foreign exchange markets each day. They want accessible trading tools that are not excessively complicated but can nonetheless be adapted to meet the specific requirements of their business with the lowest possible costs and resources.

That this is still a work in progress in one of the world’s oldest and most liquid financial markets should come as no surprise. Algos came later to the FX market than to equities and for many banks, asset management firms and corporates, they are still in the early stages of maturity. But as the FX market evolves, so too do the demands of its users. Initially they might have been willing to dabble in a new algo with fairly low expectations, but now they have more advanced requirements.

“We are having more and more conversations with our clients about algo execution strategies and how they can best apply these to their needs. They have a deep curiosity and need for a lot of detail on each strategy. Obviously, we’re very supportive in sharing these details, increasing the transparency on how our algo platform works to ensure our clients have a solid understanding from start to finish,” says Fergal Walsh, global head of algorithmic execution for foreign exchange and local markets at Citi.

“It’s important that we provide a complete package of information,” he adds. “That’s before, during and after each algo execution that clearly shows what strategy is being used, how it is progressing and how liquidity is. Having a 24/5 dedicated algo execution team that services these requests is very important to our clients – when not every detail is always available on a chart or screen.”

While it is impossible to track exactly how much of daily FX trading flows through algos today, it seems fair to assume from the number of bank and third party providers and the level of investment, that it represents a growing chunk of the market. Providers say demand for algos is rising as market participants increasingly look to reduce costs and take greater control over execution.

“By using algos, clients have an opportunity to determine the level of market impact their trading behaviour will have. Control over market impact is a major differentiator, because clients may have a very different execution horizon when compared to a traditional risk transfer,” says Kasper Folke, head of e-FX and the algo quant team at Nordea.

Using algos to increase transparency

Transparency is a key factor that attracts firms to algos as an alternative to risk transfer, particularly in an era when regulators are demanding more robust evidence of best execution. If FX market users can clearly show their stakeholders the parameters of particular algos rather than simply handing orders over to a sales contact and trusting that best execution will be achieved, they should ultimately have more satisfied investors.

Heightened demand for transparency has led to a subtle change in the way in which algos are marketed and distributed. While some of the earliest equity market algos operated under colourful names with little detail provided – based on the premise that clients were buying into the advanced technological expertise of the bank and didn’t need to see under the hood – many banks now disseminate much more granular information about how their algos work.

“The market can be clearly divided between those providers that operate very sophisticated algos with superb performance but without any real explanation of how they work, and those that really look to help their clients understand the risks they are taking and the benefits that can be achieved with particular strategies,” says Keith Hill, global co-head of e-FX sales at Societe Generale Corporate & Investment Banking (SG CIB).

The push for greater transparency and less black box algos doesn’t necessarily preclude the use of imaginative branding, as long as the reasons for that branding are clearly articulated. In addition to a hybrid time-weighted average price (TWAP+) algo, SG CIB operates two further algo strategies known as Falcon and Nightjar. Like the birds themselves, the Falcon algo is very fast while the Nightjar is stealthy and discreet. If properly explained, such names can help clients to better understand the type of strategy they are being offered.

“Explaining the meaning of the names helps clients to understand the principle behind the algo, but this is only the first step,” says Hill. “Where clients find our algo advisory to be particularly useful is when we take the time to explain exactly how the algos operate, the nature of internal and external liquidity, and the governance of the algos within the bank.”

Matching algos to trading objectives

Effective and transparent marketing is all well and good, but many buy-side firms still need support in matching their trading objectives with the most suitable and relevant algo tools. This can be a challenge because banks generally avoid giving outright trading advice or directing which tools should be used, but must rather offer constructive guidance that ensures clients have all of the relevant information they need to make their own decisions.

“Most clients will typically have a certain portion of their business that they execute on voice, algos and multi-dealer platforms, but we have to be very careful not to give explicit advice because we will never know their business as well as they do. Our priority is to be very transparent on how our five algos work so that they can make their own decisions,” says Stamos Fokianos, global head of e-business at Crédit Agricole Corporate & Investment Bank.

Arming clients with the right information to choose the algos most suited to their trading objectives can happen at a number of levels, ranging from brief high-level descriptions that very concisely explain what an algo does to much more detailed explanations of how it will perform in different market conditions and what kinds of strategy it is geared towards.

Nordea’s Folke explains that the transmission of this kind of information starts with the bank’s algo naming conventions but then extends to more advanced mechanisms to support clients through their decision-making processes.

“Our naming scheme relates closely to the trading objective, so for example the passive market-following algorithm is simply called “Make”. We have also adopted a decision tree approach that uses a series of questions to guide the client towards the algo best suited to particular trading objectives,” Folke explains.

“If you stick to treating algos as black boxes, it becomes almost impossible for clients to understand which algo might best suit their use case,” he adds. “Instead, we are trying to expose as clearly as possible the ideas behind each algorithm, which also ensures the algorithms are comparable across our entire suite.”

In many cases, the matching of algos to trading strategies comes back to the key parameters that define those algos. So in the case of SG CIB’s product set, for example, Nightjar might naturally be suited to a client with a large order that wants to avoid market impact by executing stealthily, while a trader that wants to get in and out of the market as quickly as possible would naturally opt for Falcon.

“Generally we use two axes – urgency and discretion – to map algos to execution strategies,” explains Patrick Guevel, global head of FX algo execution at SG CIB. “Nightjar is well suited to those clients that want to be very slow and discreet in their execution while Falcon is much faster. For those that are just beginning to use algos, we suggest starting with TWAP+, which has elements of both Nightjar and Falcon.”

Customisation is critical in attracting and retaining algo users.

Extending accessibility

In the initial phase of algorithmic execution, dealers would typically make their algos available to clients through their single-dealer platforms, which required firms to maintain connections to multiple platforms in order to access bank developed algos. As demand for algos expands however, they are gradually becoming more readily accessible via multi-dealer platforms and application programming interfaces (APIs).

Crédit Agricole CIB’s FX algos are still relatively new, having been launched last year, but in addition to the voice channel they are already available through several multi-dealer platforms including Bloomberg, FX Connect, 360T and FXall, and API access will be added in the coming months.

“The preferred access method depends very much on individual clients, their connectivity to the multi-dealer platforms and the strength of their relationships with salespeople. Using a platform avoids having to give out information to sales, but we have seen some very large algo orders come in on voice, often because clients like to use a salesperson that they really trust,” says Fokianos.

Direct connectivity to algos via APIs is becoming increasingly popular, Fokianos adds. “Having invested heavily in their order and execution management systems, it is much easier and more efficient for firms to send orders directly to a bank via an API to get them executed algorithmically, thereby achieving straight through processing.”

Accessibility of algos is not only about physical connections, however. It also means demystifying algo execution so that corporates and financial institutions with little experience of the practice are not intimidated by unfamiliar jargon and phraseology.

“Algo execution works best when it is delivered as a real partnership with clients so that they understand how we operate each algo and why they should use a particular strategy depending on their trading objectives,”

Corporates, for example, are increasingly using algos to improve FX execution, but some might still be put off by a lack of understanding of how different tools work. It falls to providers to ensure product names and descriptions are clear and accessible to all, and that they are providing the right level of client advisory services on a case-by-case basis.

“We try to keep the marketing as simple as possible so that clients can really understand what each algo does and how the technology works. We use machine learning and artificial intelligence to build a predictive bias into the algos, which helps both our own market makers and the clients to see where the market is going and direct the strategies accordingly,” says Jean-Michel Binefa-Kerbrat, global co-head of e-FX sales at SG CIB.

Guevel adds: “Algo execution works best when it is delivered as a real partnership with clients so that they understand how we operate each algo and why they should use a particular strategy depending on their trading objectives. It is constantly evolving and our most active clients give us ongoing feedback that helps us to develop and improve our algos.”

Leveraging TCA to manage risk

Algo execution might have tangible advantages in reducing costs and market impact but it is not without its risks, and banks say they can use transaction cost analysis (TCA) not only to satisfy regulatory requirements to pursue and evidence best execution but also to help users keep track of the risks they face.

“The big difference between traditional risk transfer and execution algos is that clients carry the market risk with algos. When choosing an algo and setting its parameters, they must weigh up the market risk against the risk of market impact. In most cases they should aim to execute as fast as possible, but not so fast that they move the market against themselves,” says Nordea’s Folke.

TCA comes in many forms and while most banks now offer some form of post-trade reporting to give clients a summary level view of how their algos performed, the drive for best execution has led to much more granular requirements for both pre- and post-trade reporting, as well as a need to combine internal and external trade analysis.

SG CIB, for example, offers its own TCA to give clients a detailed report on every trade and help them better understand liquidity as well as the trading venues that were used, but it has also partnered with TCA provider BestX to provide clients with independent validation of the quality of its algos. Analysis has shown Nightjar to perform particularly well for those clients requiring stealthy execution, says Hill.

“We have also deployed pre-trade analytics to give clients a better understanding of market liquidity and greater confidence when using algos. One of the great things about algos is that they minimise potential information leakage, which reduces execution risk at the outset. Helping clients understand which risks they are mitigating, and which risks they are taking on, in the case of an algo executing over a longer period, is a key part of our role,” Hill explains.

Nordea’s Folke agrees: “Pre-trade analytics plays a key role in ensuring the choice of algo and the parameters of that algo make sense in current market conditions, while post-trade analytics then serve to validate the pre-trade input. But analytics alone are not enough; one has to establish a feedback loop between pre- and post-trade analytics to be able to learn what is the best way to execute.”

Customising algos

Important as transparency, reporting and accessibility might be, true flexibility is only achieved when algo users are given the freedom to customise algos to suit their own individual needs. While most banks begin by offering a standard set of tools, often starting with the most basic TWAP, it is only when they grant users greater control of the algo’s parameters that they really differentiate themselves from their competitors.

“Algo strategies are often very similar from one bank to the next, but it is the customisation and controls that differ,” says Fokianos. “We have built in a lot of additional controls so that users can manage key parameters including the length of time an algo will run and its sensitivity to spot rate changes. This customisation is critical because one cannot compete by simply offering the same tools and the same liquidity as others.”

Most banks agree that customisation is critical in attracting and retaining algo users. At its most basic level, customisation might involve setting limit prices above which a client will not buy and below which a client will not sell. Those parameters can often be adjusted on the fly if market conditions necessitate more aggressive or conservative strategies.

Some banks have taken customisation a step further. “Beyond the standard parameters such as aggressiveness and execution time, we offer customisation that goes further under the hood of the algos, such as liquidity sources and algorithm logic, but we have not yet seen much interest in this,” says Nordea’s Folke. “While some clients do have very specialised requirements that require customisation, we feel mostly that the standard parameters of the algos offer sufficient choice to cover the needs of our clients.”

Direct connectivity to algos via APIs is becoming increasingly popular

Customisation is not only the domain of the banks, however. Some multi-dealer platforms are taking the demand for greater control to the next level by allowing users to switch from one algo to another while trades are already in progress. This switching functionality is not yet widely available, but it offers a glimpse of the kind of advanced user control towards which the FX algo world is slowly moving.

“The greatest risk when using an algo is managing sudden market moves, and some multi-dealer platforms will now allow users to switch algos during execution. This is a very advanced feature that means customisation is no longer confined to fine tuning, but one can actually switch from a passive to an aggressive algo or vice versa, thereby managing risk on the fly,” says Fokianos.

Ultimately the true test of an algo is whether it guides the user to the best liquidity in a particular currency pair in the most efficient way. As banks continue to develop their algo execution services, customisation, flexibility and transparency will remain critical, but it is the access to liquidity that will likely determine where firms choose to source their algos on an ongoing basis.

“We have invested in ensuring we use those trading venues where there is deep liquidity available in a very efficient and low-footprint way. As FX liquidity has fragmented across multiple trading venues, placing order interest and minimising market impact has become a more skilled job, which has strengthened the case for algo execution,” says Citi’s Walsh.