There is now a new landscape when it comes to how FX algo clients engage with algo providers, with clients increasingly showing a better understanding of how the algos work, says Ralf Donner, Head of FICC Execution Solutions at Goldman Sachs. “This translates into the kind of questions which clients ask us, which can be quite detailed in some cases,” he adds. “It’s common when onboarding a new client today for them to not just be interested in hearing a description of our algos but increasingly to also be interested in looking at the analytics. They might want to see past results or they might want to see how the algo has performed historically and in different currency pairs, different size buckets, or they might want to know about how we use various venues for execution, how we use our internal liquidity, how we compete orders etc. There’s a growing understanding among clients about how the algos work which leads to them being more interested in having these detailed discussions with us.”
Clients have increasingly started comparing and contrasting the performance of algo strategies between various providers and asking about whether an algo boasts a specific feature. According to Donner, those comparisons between the products offered by different providers can also be useful to the bank, as that provides additional intelligence about what is on offer today which may or may not be useful in the provider’s own products. It is also useful to see what concerns a client may have about an algo product, he adds. The team already gets feedback from internal user of the algo products within the bank, but it is important to also hear external feedback from clients as their requirements might be different to those of the trading desks or other internal users.
“This can be very useful and some of our products have even been inspired by client discussions,” Donner says. “The basket algo, for instance, which is one of our most sophisticated products, was inspired by a conversation with a hedge fund in the United States. As a result of that conversation, we suddenly realised that there was a broader need for such a product. So on occasion, we might even have a client conversation on the inner workings of an algo that leads to a whole new product.”
At Commerzbank, having these types of discussions has also helped to deepen the relationship with partner clients, says Nickolas Congdon, Director – Head of eTrading Services at Commerzbank. “Our algo strategies for clients are derived from the same proprietary algo strategies and analytical technology as those used to electronically risk manage our global forex business. We have taken these core components, coupled with client feedback and delivered a set of adaptive algo strategies and execution tools that supports clients in exceeding their trading objectives. Offering FX algos to clients was really just an extension of this, allowing us to bring the components together with clients and what they required, understanding that this may be something they may have not been as familiar with previously”.
The aim, really, is education, Congdon says. It is about helping clients understand which tool is best suited for each given scenario and this is a very fluid process, he explains. “Every scenario will naturally be different because clients will have different underlying execution objectives and varying levels of comfort using the tools at their disposal. Our job is to create a level of comfort by removing the barriers and providing a common ground irrespective of client sophistication. This is the basis of Principle 18 in the FX Global Code. This also benefits the algo provider,” Congdon adds. “Whenever we can provide a service to clients, such as helping with their algo selection or a TCA report, this deepens the relationship as their partner bank.”
“It is about knowing the right questions to ask your liquidity and algo provider. Not just when clients are executing with us, but with all their counterparties. Our clients know that they can rely on us to provide that level of transparency and they find comfort in that. So, when they have a larger amount to execute they know they can reach out to us and we’ll help them to manage that order, explain which strategy to utilise for certain currency pairs and how their order interacts with multiple pools of liquidity globally across time zones,” says Congdon.
The way the FX algo platform operates at BNP Paribas means that nine times out of ten, one of the existing three flavours of algorithm will already meet the needs of a particular client, Asif Razaq, Global Head of FX Automated Client Execution at BNP Paribas, says. He explains that the reason why that works is because the bank has engineered adaptive algorithms, so the algorithms are not a fixed style of execution and they can change their behaviour to fit the market.
Elements of Control
On top of this, BNP Paribas also added the concept of interactive elements to the algo strategies during its fourth generation roll out, allowing clients to also exercise a certain degree of control over the flight path of the algo by changing some parameters in flight. “Occasionally, we have a client who actually wants to do something very different or very bespoke with their algo execution,” says Razaq. “That’s why we introduced an additional service called Flex, which allows us to create a bespoke algo strategy for the client based on their specific execution requirements. This is very important for us because this basically then gives us the ability to provide a very bespoke customised service for our clients, meeting their needs exactly.” This service has proved to be increasingly popular since its launch, Razaq notes, and is very important as, when done successfully, it cultivates more flow coming into BNP’s algo platform. It is also a service which requires his team to work quite closely with clients and, in certain cases, requires the team to be very careful when explaining the potential impact of something that the clients are looking to achieve, Razaq says. He adds: “Sometimes we simply have to say no. We have to explain to some clients that we cannot offer what they are asking for because the type of behaviour they are hoping to exhibit in an algorithm could be too disruptive for market. But Flex has now become a core part of our platform due to the number of experienced algo users asking for very specific behaviours or very specific functionality from the core set of algo strategies.”
In addition, Razaq says that when it comes to showing an increased interest in how algo toolsets work in very specific areas, it tends to be a very specific segment of the buyside community that want to delve down that path. He explains that some clients just use the off the shelf algorithms to ‘fire and forget’, so are fairly hands off during the execution. “The second group, who we call active execution managers or traders, will be very hands on and might change the flight path of the algo mid execution,” Razaq adds. “This is where clients would interact with our innovative digital trading assistant, ALiX, who becomes the voice of the algo, providing real time commentary to the client, giving them feedback on what they’re seeing in the market. That’s where you see a symbiotic relationship between man and machine, working together with the tools provided to execute in the market.” But then there is also a third camp, Razaq continues, who want the algo to behave in a very specific way based on their requirements.
Too Much Transparency?
“This kind of client wants to look at the inner details of the algo. This is why we started introducing Flex to that client base, which tends to consist of more systematic, high frequency execution clients,” Razaq says. “This group is interesting because they are often clients that had previously tended to build their own algorithms in house, which they would deploy to then go and source liquidity in FX through their bank counterparts. But what’s happening now is that this group has seen how the FX algo market has evolved over the past three years among the other client groups and they are now re-evaluating whether such algos can also be tweaked to fit their execution requirements as well.” He adds that there are additional benefits to outsourcing, such as BNP’s very significant resource pool which provides maintenance, coverage, tuning, developing new algos or offering services like Flex, whereas a buyside firm may not have that agility or those resources.
Furthermore, in order to conduct analysis of the performance of an execution algorithm, client’s also need data – and lots of it, says Razaq. “So if you’re a client and you’re only executing 20 trades a day, then your dataset is going to be very limited. At BNP, we have a much larger dataset to use when evaluating the performance of our algos, meaning the output is far richer, we have more data points and so we are able to provide more calculated and more quantitative solutions for improving our algo strategies compared to a clients limited data set.” There might be a point where client involvement in the FX algo process might tip into negative territory.
One area where this can happen is when a client has a dataset and believes it points to certain conclusion, says Donner, but in reality they are only looking at a fraction of the complete dataset that the bank has access to. As a result, it can be possible for a client to arrive at a slightly incorrect conclusion because they haven’t taken a look at a large enough dataset, he warns.
“Take mid books, for example, which are generally acknowledged to be beneficial to algo executions because they help to offset risk at mid with peers and do so very quietly,” Donner says. “But can be some clients who are a little wary of mid books because they’ve heard that they might incur penalty charges if they were to overuse a mid book.” It is true that pummelling the mid book with very aggressive trades would have negative consequences for algo performance, but that is all baked into the algo logic already,” Donner explains. “So it is sometimes possible for a client to come to what we would consider a slightly over anxious conclusion about something, such as asking us to remove mid books from the algo. That would not be beneficial to client in terms of best execution,” he adds.
Bringing Clarity to Complexity
On the other hand, the growth of analytics, both in house and third party, is also driving a desire among clients to understand how the algo works, says Donner. “On the post trade side, sometimes we will even join in a discussion with the analytics provider and the client together to discuss performance, which can be an interesting discussion for all because the analytics provider may have additional insights into what to look out for,” he adds. “The other aspect is that we now have a very rich dataset for pre trade analytics as well. So clients ask us a lot of questions around this, such as interesting phenomena that they see in patterns of liquidity, what that means for choice of algo and during execution they’re of course, looking in detail at the nature of the fills, whether it’s passive or aggressive and whether there is any market impact observable from execution.” Overall, increased transparency just serves to strengthens the relationship between client and algo provider, Donner explains. “It adds another relationship that the client has with the bank – for example, they might build a strong relationship with an algo quant team to discuss some of the technical details,” he adds. “This provides clients with another access point that they have to the sell side and that differs a bit from client to client.”
Not one size fits all
The level of involvement in the FX algo process by clients also needs to take into account a number of different variables, which the algo provider can help them with on a case by a case basis, says Congdon. “It’s definitely not a one size fits all,” he explains. “They may want to execute a large order late on a Friday afternoon, but that may not be the best time to do it given thinning liquidity conditions. Have they considered the current macro environment? What is the user’s product understanding and risk appetite? The client is in the driving seat at the end of the day, but our proprietary analysis will show an execution profile and we can make a recommendation based on a holistic data set. Ultimately it’s up to them, our goal is to provide them with unambiguous information so they can make the best decision based on their execution goals.”
The most important thing is that the client understands what they are trying to do, which is the aim of the new measures which were recommended in the most recent FX Global Code, Congdon adds. “The Code has created an industry standard questionnaire for use by banks, on a voluntary basis, in a bid to demystify all of the fancy names and parameters that the banks have for their algos,” he says. “This should help the clients to compare and contrast the FX algos the are available, apples to apples.” The Algo Due Diligence Questionnaire should provide clients with a meaningful understanding of the underlying strategies, any potential conflicts and the role the provider plays while arming less experienced clients with information they may not have gained otherwise, Congdon explains.
Catering to requirements
“The goal is to increase that transparency and make sure the information available is at a level that everyone can understand,” he adds. “The level of awareness within the buyside still varies significantly.” For some, FX is viewed as an unwanted consequence of fundamental business and is not core for them, while at the other end of the spectrum we see highly sophisticated financial institutions who have requested bespoke algorithm strategies, Congdon says. “But across the whole spectrum there is that common theme, the underlying concept that they can expect increased execution transparency and heightened awareness of the reduced costs this order type provides,” he adds.
Clients do need to have a high level of understanding about the nature of the execution algorithm and what it is going to do, but in terms of understanding the inner workings of the technology, that’s when it becomes a bit too tricky, warns Razaq. “There is also an issue of intellectual property,” he adds. “We need to protect the IP of BNP’s algo suite, so there’s only a certain amount of information we can share. There is a level of detail that we are willing to go into with clients to understand how the algorithms are functioning and what their goals are, but there’s only a limited level of depth we can go into without giving away the secret recipe of the algo.”
Transparency has been a core of the algo product since day one, Razaq adds. “When BNP launched its first FX algos some 10 years ago, the key benefit we were explaining to clients was that by using the algo, they would for the first time get full transparency of their execution. It would mean they would be able to get a post trade report detailing every single child order to the nearest millisecond, where it had sourced that liquidity from and at what price. It was the first time that clients had that level of transparency and it was because of the TCA element that we built into the strategy.”
The algos also give clients the confidence that the analytics document fully auditable, it will provide all the data they might need to validate their one trade, they get performance benchmarks. “Clients are now also getting transparency not only on the performance of the algorithm, but on how the algo works, how we are set up, if we are segregated form the trading desk – the flow may be very sensitive and they don’t want the trading desk to be aware that they are trading the market at that size,” Razaq concludes. “Clients now expect that 360 degree perspective when it comes to algo perfomance, whereas before it was very much straight down the line. Transparency has always been a big feature of the algos and a big draw factor for clients. This is just taking it to another level.”