Further trends highlighted by the report: FX Trading: Strategic Importance of Electronification and Automation, included access to better data and analytics, which was mentioned by 30% of respondents. “The onus then is on the end user’s firm to implement guardrails to prevent poor execution, whether algo- or human-driven. This explains the investment focus on execution management and analytics; getting an uncompetitive price is a high risk and could lead to negative financial consequences,” says Stephen Bruel, Senior Analyst at Coalition Greenwich.
Gurpreet Ubhi, EMEA Head of eFX Sales at Morgan Stanley, agrees, noting that in terms of customisation, FX algo clients are increasingly looking for ways to add flexibility while they are running orders. “We offer a number of different ways for them to achieve this,” he adds. “It is important to allow flexibility to vary the trading speed of the execution, but we also have the ability to choose to trade with momentum or reversion, or even switching strategy completely within an inflight order. Not every client is the same, so having the ability to adjust these parameters allows clients to be able to adjust an algo based on their specific trading style or execution needs.”
“One of the appeals of algos is that they can be low to no touch execution and clients want to systemise this.”
Gurpreet Ubhi
Ubhi explains that the FX algo offering at Morgan Stanley leverages the strength of their existing equities infrastructure and the market leading liquidity and analytics platform Quantitative Solutions and Innovations (QSI), which is heavily utilised by clients. In addition, the bank curates FX liquidity pools to allow clients to leverage internalisation. “One important feature for most algos is their ability to trade with an internal matching pool and open up more potential liquidity through non-market visible sources. However, different internalization methodologies exist, with some supporting more opportunities for bid-offer spread capture than others” says Ubhi. “We pride ourselves on our ability to reduce market impact through monitoring liquidity pool profiles. In particular, our analysis shows that executions against an internal pool or internal streams can have benefits in terms of market footprint and decay profile, so it makes sense to maximise that usage wherever possible.”
Balancing market impact and risk
According to Ubhi, it is worth comparing internalisation along with every other liquidity source when it comes to market impact. He explains: “It is easy to state that internalization must be better, but we believe it is still important for any provider to consider whether the child slices that are ‘internalized’ are truly that, and are not simply back-to-backed to the market. In our case, this applies equally to the flow through our client matching pool and also any trading against our principal liquidity stream. Breaking out this kind of nuance on the TCA report provides transparency for clients.”
Mark Rendel, Quant Executive Director at Morgan Stanley, adds that on the FX side, the bank is increasingly focused on trying to pull out market signals as a way to decrease the cost for the client.
“We see our client relationships as true partnerships, working together in some cases for a number of years to continuously optimise their algo execution performance.”
Mark Rendel
He adds that “We offer a lot of flexibility to those clients who want it. However we are conscious that there is always the need to strike a balance between flexibility and complexity to avoid the risk of overwhelming the user with too many obscure settings and parameters.” There is also no ‘one-size-fits-all’ when developing functionality for algo clients, he explains. “Some clients want every tool and functionality available, and others want to leverage the fact that we have access to a large data set,” says Rendel. “We can do the analysis for them, try and work out what works best for their flow specifically, and make recommendations based on that, or work with them to make those changes on their behalf with their approval.”
Focus on transparency
In addition, the increasing demand for bespoke customisation tends to stem from more sophisticated clients, says Dr John Quayle, Head of Client Algo Execution at NatWest Markets. As a result, liquidity providers need to tailor a given algo for the client in question, ensuring that all specific client settings are automatically set to a default for each algo type and currency pair going forwards, he explains, while still leaving open the option for the client to amend the execution in real-time. “In particular, being able to change the liquidity pool selection on a real-time basis is proving to be highly valued by clients,” Quayle says. “This is available in our Peg Clipper algo now and is probably the most effective way to adjust behaviour to suit the prevailing market conditions; it allows for quicker executions without the need to place child orders more aggressively inside the top-of-book spread, which can risk undue market impact,” he warns.
The ability for clients to have 100% visibility on all aspects of an execution has always been one of the selling points of FX algos, Quayle continues. Aggregated performance metrics from third-party TCA providers has enhanced this, he explains, alongside pre-trade metrics such as likely slippage and fill times. He adds: “The sell-side ought to be providing comprehensive details around how the algo works as well. Clients should have confidence that algos are repeatable and reliable and will never engage in any ‘black-box’ behaviour which cannot be explained after the fact.”
Furthermore, Quayle notes that some client types often do not want to have interact with the algo, preferring instead to use the pre-set parameters or tailored settings offered by the algo provider. “Some clients do not have the time to try remembering specific settings themselves for a variety of different bank algos. Instead, the algo needs to be able to support a range of requests for different types of execution or trading behaviours from different clients,” he adds. This speaks to flexibility in a specific algo, says Quayle.
For example, providers need to offer algos which cover the speed spectrum from quick to slow, but the algo user needs to remember which to select. “Again, some clients do not want to do this – they would rather have one algo and have a very clear dial to speed up the execution or slow it down. So being able to provide easily accessible flexibility is important. It is a slight contradiction in terms – but the way we achieve this in the Peg Clipper is to offer a flexible algo which can default on a counterparty basis in quite a granular manner,” Quayle says. In addition, there is a pipeline of new venues and clients are asking their LPs to provide access to these venues via algos, he explains. “There is increasing recognition among the clients we talk to that the liquidity the algo accesses is as important as the algo itself,” adds Quayle.
“In particular, being able to change the liquidity pool selection on a real-time basis is proving to be highly valued by clients”
Dr John Quayle
Increased flexibility and choice
Over the past few years there has been a trend away from increasingly complicated ‘black box’ style algos, towards ones which are more flexible in-flight, agrees Aled Basey, Head of HausFX & FX Execution Advisory, UK at Deutsche Bank. He notes that this trend has in part been shaped by regulation, with the senior managers regime requiring heads of trading to be able to explain the products they are using. “Catering to this demand means real time analytics, delivered both pre-trade and in flight, are now a prerequisite for many clients,” Basey says. “Different clients use algos for different reasons, and equally different algos are designed to target different results. The correct starting point is ensuring that the client and the bank are aligned on their primary objective. Our execution advisory team collaborate with clients to help them pick the right tools for the job, which in turn improves the user experience and probability of outperformance.”
Basey adds that FX algo trading is arguably the most transparent form of execution in the market. Clients are now able to see every underlying fill both in real time and post-trade, which can then be verified with Deutsche Bank’s proprietary class leading Market Colour TCA app, or through its partners, BestX and Tradefeedr, he explains. “With more data comes the ability to help clients make more informed decisions, but the real value add is digesting these huge data sets down into simple, actionable, decision enhancing insights. A great example of that is the recent addition of our ‘Quick Pre Trade’ feature in AutobahnFX, which provides indications of duration and probability weighted outcomes across different urgency settings,” says Basey.
“The next wave of analytics within Deutsche Bank are solving for ‘when’ to use an algo, which cannot be done without factoring in tradable risk prices.”
Aled Basey
According to Basey, the market has also now passed the ‘high water’ mark of different variations of algo strategies. “The focus has turned instead to optimising the user experience and providing the best of both across risk transfer and algos,” he says. “To this end, we have recently released a new Instant Market Access ‘IMA’ feature, which allows clients to launch their preferred algo strategy in one click. This has proved incredibly popular among our clients during recent periods of heightened volatility, saving them time on the configuration and instead focusing on adding their alpha through the in-flight management.”
Advanced workflow solutions
Rendel adds that one area where Morgan Stanley specialises in is forging a true partnership with clients. “Knowing what the client’s expectations and desired outcomes are, allows us to do analysis with them. We can combine that information with our deep knowledge of how each individual algo, parameter or liquidity source works to optimize the best settings to achieve those aims. It also allows us to leverage our performance statistics as a whole to improve performance for any one specific client. We see our client relationships as true partnerships, working together in some cases for a number of years to continuously optimise their algo execution performance.” Ubhi agrees, adding that the bank also provides clients the flexibility to adjust liquidity venues within certain order types. “For example some clients may prefer to only execute against non-primary market venues” he explains. “At the same time, others prefer the bank to do the configuration for them.”
Ubhi notes that the market analytics have become more sophisticated over the last few years, and has helped provide market transparency to clients around liquidity conditions, allowing them to make more informed execution decisions. In addition, Ubhi believes that the market has seen a trend towards automated algo wheels, with certain algo clients increasingly adopting an equity-style framework. “The landscape has developed in that regard because more of the technology providers can see there is client demand for this type of workflow. It can remove some of the discretionary bias that might exist at present” Ubhi says. “It also allows clients to experiment in a rigorous and systematic way with new order types that they may not have been using regularly before. One of the appeals of algos is that they can be low to no touch execution and clients want to systemise this. They want to send orders without having to be overly involved in the selection, so they really want to base their rotation on the performance of the strategies.”
He continues: “We are happy to have the dialogue with our clients to help them adopt specific settings that align with their objectives i.e. if a client is more interested in limiting execution risk or minimizing average execution cost, or anything in between. We are finding clients who typically execute over longer periods of the day, might have an objective to follow a specific volume or time based benchmark, which requires a different set of considerations.
Enhanced decision making
“Some clients are more interested in the micro level details of order placement. Realtime TCA has become increasingly important to facilitate this level of transparency and slice level breakdowns” Ubhi says. “We do still find that in some cases it makes more sense to aggregate order level analysis to pull out patterns for our clients” In addition, Morgan Stanley offers its QSI data services not just through Matrix®, but also via API, notes Rendel. “This allows clients to query pre-trade cost estimates for an execution, and include that information in the client-side decision making process” he explains.
Ultimately, the decision whether or not to do a risk-transfer vs say a passive algo – quick or slow execution – is very situation dependent, says Quayle. He explains that there is value in certainty – or a cost to uncertainty – which may well make risk transfer more desirable in some circumstances, but this will usually come at the expense of crossing more spread on average than using an algo. “The reason algos exist is that they will save money on a long-term basis,” Quayle adds. “These two scenarios also represent two extremes – and between these there is a wide range of possible approaches. Filling in this gap to give full access to the spectrum of possible choices is something that we strive to achieve at NatWest. Users can now either tailor underlying parameters, such as level of aggression or liquidity pool breadth themselves, or simply set a target duration and the algo will optimise the parameter itself in a fully automated manner to achieve the target outcome.” Additionally, being open to, and able to onboard new venues quickly is also valued by clients, notes Quayle. He adds: “There is a pipeline of innovative new trading venues and whilst not all will succeed, giving clients the option to participate and the control to choose how and when to use these venues is highly valued.”
There are also two further key areas where LPs can add value, according to Quayle. The first is in providing genuinely useful ‘real time’ TCA metrics directly to the client, so that the client can be better informed to make any changes to the algo setting during its operation, he explains. The second, and which Quayle believes has far greater potential, is to be able to implement an automated A/B testing framework customised to the client. “Being able to test the performance of different parameter sets is of course possible – the client can alternate between successive algos for example, but the rate of data collection is slow and manual. A fully automated solution which can toggle between parameter sets on a higher frequency basis automatically, and aggregate the data to determine which performs better, would be extremely helpful to the client and this is what we are building at NatWest for our more active clients,” he says.
Consultancy and collaboration
Overall, Quayle believes that the best outcomes for algo users and providers can only be achieved through a very strong cooperative relationship and a dynamic feedback loop between the user and LP. “When this happens, the LP has a detailed understanding of the requirements, is able to make prompt and innovative improvements to the algo logic and work with the user to collect and analyse relevant performance data,” says Quayle. “At NatWest, we have a clear focus on getting the best outcome for clients; developing an environment of continuous improvement is key to consistently achieving this.”
This area of the market is developing rapidly, notes Basey, with clients looking to save more time than simply in the final mile of the execution. “Our HausFX team specialise in partnering with clients to transform their FX trade lifecycle, reducing cost and operational risk whilst freeing up time for alpha generating tasks,” he adds. In addition, Basey says that having witnessed both sides of the independent vs bank TCA debate, he believes there are ultimately different use cases for both. Independent post trade is now well commoditised and great for evidencing best execution purposes, broker reviews and comparing between providers across a full book of business, Basey adds, yet there are challenges with attempting to utilise this data at the point of trade. “This requires both real-time analysis and adjusting results for similar market conditions and the more you slice the data, the smaller and less relevant the underlying sample size,” he warns.
“By contrast, bank-own analytics have the edge in both pre- and in-flight analytics, as the technology stack already run co-location services and sit upon a plethora of real-time market and transaction data,” Basey continues. “This provides a ‘franchise’ view that the independent providers would love to have but are ultimately constrained by their respective client bases.” This also allows banks like Deutsche Bank to develop ‘market regime’ metrics which again help provide clients guidance in-flight to improve the probability of achieving their objective, explains Basey. He adds: “The next wave of analytics within Deutsche Bank are solving for ‘when’ to use an algo, which cannot be done without factoring in tradable risk prices – this sits firmly within the Bank’s area of expertise.” According to Basey, algo execution is by default agency execution – and any agency relationship requires trust, transparency and an alignment of objectives. “We help to solve our client’s FX problems, whether that is via FX algos or using broader workflow solutions. We tackle both of these head on within the Execution Advisory and HausFX teams. At Deutsche Bank, we believe in partnering and really utilising the data to help clients to make the right choices for their underlying investors,” he concludes.
