GlaxoSmithKline (GSK) is one of the world’s most innovative, best performing and trusted healthcare companies. It has three global businesses that research, develop and manufacture innovative pharmaceutical medicines, vaccines and consumer healthcare products. FXAlgoNews spoke with Fabio Piga, MSTA, Group Treasury Front Office Manager - FX & Interest Rate Risk Management at the firm about algorithmic FX trading and the key benefits it delivers for GSK.
Fabio, please tell us a little about your day to day responsibilities at GSK and what your job involves.
I joined GSK 2 years ago as treasury front office manager and I am responsible for all risk management execution activities relating to cash management, foreign exchange and interest rate risk, and ensuring the delivery of a high-quality front office function to the business. I work closely with various areas of the business to support payables, receivables and inventory management and enhancement of the Treasury Management System (TMS) and execution processes.
How does your FX dealing unit fit into GSK’s overall treasury operations?
The FX dealing team is at the core of Treasury operations; we are the SMEs for FX and manage the organisation’s liquidity and financial risks, banking relationships and working capital as well as providing support to management and business units. We are also involved in the FX and interest rate risk management activities associated with M&A.
How would you describe the key objectives and guiding principles of your trading desk?
Our key objective is to make sure that FX risk is identified and managed in a timely and efficient way, in line with GSK’s FX policies. We cover the full FX exposures as soon as they arise, and we make sure there are no open risks for GSK. We will start soon hedging more forecast flows and introducing new execution strategies and hedging programmes by pairs.
Some firms feel it’s important to use a stable and mature platform rather than the latest technology. How much importance does GSK place on increasing automation and continuously working to improve its own trading infrastructure?
GSK treasury is making significant changes to its use of technology. It is in the process of replacing its TMS and FX execution platform at the same time. We are moving to system that allows more automation and gives the team more time to analyse the market and focus on greater cost reduction. I am personally very keen on increasing automation and reducing manual interaction with the orders being executed, so that the traders can spend more time on tasks that add value i.e. reducing execution costs, TCA.
Did GSK suddenly decide to adopt algorithmic FX trading or was it a more gradual process?
I would say it was a quite fast process. I introduced algo trading in my first 4 months here. I spent those initial months analysing GSK currency portfolio, execution style and processes in order to gather a better understanding of how the company operates.
Making changes in a big organisation is always a challenge so it’s really important to involve the right people in the conversation at the right time. I had to make sure a switch to algo trading would not be disruptive for the team, the system or the existing processes.
In the end, it was a success and senior management has been very supportive and happy with the change.
What are your main goals when undertaking algorithmic FX trading and what types of FX orders are usually a good fit for it?
The main goal of GSK algo execution is to reduce execution costs by capturing spread and achieving cost savings versus risk transfer whilst minimising market impact. When we hedge, we don’t express a view on the currency, and we don’t normally have a specific target price to beat. Some of the other benefits of introducing algos included the efficiency versus traditional voice execution, and transparent TCA.
In terms of the “types of FX orders are usually a good fit”, I believe the FX portfolio drives the strategy and not vice versa. We adapted/designed the strategy around our FX exposures and company needs. More than 80% of what we trade is G10, hence very liquid pairs that can be traded using a passive strategy.
Post-trade TCA has great value of course but how important is the ability for you to see how an algorithm is performing in real-time and how can that be achieved?
I believe that real-time TCA is essential, particularly for orders with a longer time horizon, which are most likely to be impacted by changing market conditions.
Real time market data such as TOB spreads, liquidity, execution venues, economic data releases can be used for a better decision-making process and to adjust the strategy under changing market conditions. Few algo providers have invested heavily in real time TCA and we started using them more and more, especially those that prioritized ease-of-access via Bloomberg. What is important is making sure you know your algo before interacting with it; most third generation adaptive algos are actually designed to read markets in real time, understand the changes and then formulate dynamic strategy accordingly. The question is do you think you can do better than a machine?
In what ways can improved TCA help you to make more effective use of FX algos?
TCA is key for us to understand algo performance and our performance too.We use a combination of pre trade, real time and post trade TCA to make trading decisions and this has proved to be beneficial and improved our execution costs.
Initially we used the post trade TCA only, but then started adding pre trade analysis to get a better view of liquidity, volumes before trading and adjust to changes whilst the order is executing with real time tools.
I found that post trade TCAs biggest problem is the lack of a centralised FX market benchmark and pre trade TCA looking at a historical market does not work well in “unusual scenarios”.
What types of FX algorithms are your trading team currently employing and what factors generally influence this?
We tend to be extremely passive with our algorithmic orders, opting for less impact and more spread savings. This decision was made after analysing our currency portfolio which is mainly made of very liquid pairs (85% G10) . Of course, the type of FX algorithm a trader decides to use depends on benchmark, order size, currency pair, liquidity, and urgency of the FX order. We want to be dynamic and able to adapt execution if market condition changes. An example has been the COVID situation; on some days we realised that passive algos were not performing as expected due to the poor liquidity and high volatility hence we chose a more aggressive type of execution, in order to get the order done and close out the risk.
How willing are you to let algos do their job without micromanaging many aspects of the execution process?
I think that as a trader you are ultimately responsible for your trade performance and shouldn’t rely 100% on your algo to do the right thing. This is true especially when market conditions are not normal; this is where I believe the trader can still add some value vs the algo.
Using real time TCA can be a good starting point to understand how your algo is performing and when it’s not available, I like interacting with the algo teams over chat to have a better understanding of market conditions and make sure the strategy I’m using will deliver the results I’m expecting.
What do you see as the key benefits that algorithmic FX trading delivers for GSK?
The main benefit of introducing algorithmic FX trading has been the reduction of execution costs vs mid, which are significant. Not only do we not pay the premium of the risk transfer price, we also save against the mid-price.
Algo trading makes even more sense now because banks no longer warehouse risk for very long. Other benefits have been the access to a wider liquidity pool, full control over execution, transparency of execution with detailed TCA, reduction of time spent on a deal and the ability to analyse large amounts of data in order to conduct price discovery and identify optimal liquidity.
Have your algos performed as expected during the recent Covid-19 crisis and have they proved resilient under sometimes quite volatile conditions?
The most significant change that we’ve seen as a result of the current market conditions is the fact that passive strategy execution has been underperforming. With greater market volatility, wider spreads, less liquidity beyond top of the book, executing via a pure passive strategy has proved challenging.
We noticed a big difference on spreads in the primary vs secondary market and that even passive orders become ‘impactful’ as they feed back into moves in these markets.
As a result, we adjusted our execution moving toward a TWAP style algo, which allowed us to get risk covered even if crossing spread whilst still taking passive fills opportunities. Furthermore, to reduce market impact we routed our trades to firm liquidity venues and primary markets only.
In what ways has your TCA been influenced by market conditions during this event and have you been able to pick up any important lessons for your future algo trading activities from it?
On the TCA side, we increased usage of pre trade TCA tools to get a better view of volatility, spreads and select the most optimal execution type: Algo vs RT vs Streaming. When executing via algo we leveraged in-flight TCA to direct flows to specific venues or reduce/increase the execution speed, so we are interacting more with the algo than we did before.
In these market conditions, passive algos have not performed as they did in the past so we have learnt how to adapt and be flexible. For example, if the market is moving against you and you passive algo is not getting fills, we have been willing to pay the spread to close our risk sooner.
What steps do you think banks and algo providers can take to increase the appeal of algorithmic execution especially for large corporates?
I think banks need to reduce the complexity of their algo offering for corporates and improve understanding of internal processes and strategies. Corporates are very fragmented when it comes to FX execution, some have a very strict policy, which is difficult to change, others see FX dealing as an admin task and don’t focus/invest in it, whilst others instead are very sophisticated and behave like real money clients.
In such environments, banks need to approach clients differently: guiding, training and supporting those who are behind the curve and working closely with the more sophisticated ones to make sure they are always informed and included in changes.
We are now frequently talking about how AI and Machine Learning are being leveraged in algorithmic FX trading and associated analytics. How are you approaching the use of next generation technology like this to improve your own FX trading outcomes?
We are not considering it at the moment, but I agree that with the implementation of AI for FX Trading, computers will undoubtedly make accurate and precise decisions saving us time and energy.
In my opinion it will be very challenging to figure out what kind of data and data combinations would be the most appropriate when building an FX trading model and AI will still face the challenges of filtering out the unpredictability of the markets.For AI to reach the required level of efficiency in FX, it needs the cooperation of experts and experienced FX traders.
Many large asset managers are already adopting algorithmic FX trading but most corporates haven’t yet fully committed to it. Do you expect that to change?
Yes definitely. Corporates want to minimize FX costs in the same way as other buy-side clients and are unwilling to pay a widespread via RFQ.
Some recent studies showed that corporates are looking to execute more FX spot volume algorithmically and are increasingly using bank-provided algos, attracted by their potential benefits like cost savings through minimizing spreads paid, reduce market impact (especially for larger trades), access to liquidity across a wide range of liquidity pools.
Corporate FX desks can also benefit from desk operational efficiencies that result from automating the execution.
In what ways are you likely to expand your use of FX algos in the future?
I would like to be able to increase the percentage of spot executed via algos. We’re currently being stopped by system limitations, many of which will be fixed with the launch of our new TMS. The new system will allow us to net and aggregate our exposure minimizing spread crossing.
In addition, I would like the team to increase their confidence in interacting with the algo and be more dynamic in adjusting algo selection vs market conditions.
I believe that the automation and time we are saving on execution thanks to algos should be spent on TCA analysis and data analysis to improve the quality of our execution and add value for GSK stakeholders.