Andrew, you are a hugely experienced treasury trading practitioner. Please tell us a little about what your job at Shell typically involves.
My team is responsible for the execution of all financial market instruments on behalf of central treasury, which includes FX spot, forwards and swaps. As a business we still deal in largely USD priced energy, but we have very large revenues and costs in various regions and therefore other currencies. As a result we trade significant volume in a range of pairs daily.
We also execute FX hedges for our commodity trading businesses, which can be both significant in size and very volatile due to movements in the underlying contracts.
Finally, we execute on behalf of our local businesses in various onshore and more regulated markets.
How would you describe the key objectives and guiding principles of your FX dealing activities and market execution strategies?
Put simply, we try to maximise the USD value of the dealing we are required to do. As a first step we try to minimise our go-to-market flow by offsetting various internal exposures wherever possible.
With our residual exposures, we don’t attempt to call market direction; rather we look to minimise the cost of execution by achieving tight spreads and reducing any signalling or information leakage that may move markets in an unfavourable direction.
We do this by accessing the greatest variety of liquidity and by generally dealing in a passive manner, while avoiding where possible periods of high volatility, such as market data releases.
Shell was very much a pioneer in utilising FX execution algorithms. How important have they now become in your day to day trading operations?
Algos make up a significant proportion of our spot volumes, with multi dealer ‘in competition’ tickets making up the rest. The overall volume of algo flow depends on the size and nature of the dealing required on any one day.
What are your main objectives when undertaking algorithmic FX trading and what types of orders are usually a good fit for them?
As a corporate we can perhaps be more patient with our execution than some other market participants. Algos can then help us to act as a more passive liquidity taker rather than simply paying a bank to take that risk from us.
This can be especially beneficial if we have very large size regular flow, or if we are executing in a less liquid pair where the spread cost of simply passing our risk to a bank often outweighs the risk to ourselves of holding for a longer period of time.
On the other hand, different algo types can also be effective when we have time-sensitive flow and speed is a greater requirement. While more ‘aggressive’ execution may be required I find the impact tends to be less than simply risk transferring.
How do you source your FX algos and what issues influence how you go about that?
All of our FX execution, including algo access, is currently via Bloomberg FXGO. We share our algo flow around our large group of relationship banking counterparts.
However, the nuances between algo providers in terms of speed of execution, the internalisation or matching capabilities, as well as the ability to adjust settings in flight are all considerations when choosing the algo provider and type.
How much real-time visibility do you seek on how an algo is performing during the execution process?
Real time data can be a useful addition to the service and several of our banks now do provide this as standard. It gives comfort that the order is executing as would be expected and can serve to highlight when settings, such as speed, can be updated. I find it particularly useful with less liquid currencies and NDFs.
As a business we still deal in largely USD priced energy, but we have very large revenues and costs in various regions and therefore other currencies. As a result we trade significant volume in a range of pairs daily.
How do you analyse the results of your algorithmic FX trading to see how effective it is and what metrics are useful for achieving this?
We use a 3rd party TCA to calculate and analyse spread savings and fill compared to the arrival mid and then we take a holistic view of the performance. As we have a generally passive approach, we are willing to accept a wider deviation from the arrival rate to reduce spread paid, but we do also monitor closely to ensure this is not a consistent miss.
What other data related to algorithmic FX trading are you trying to capture which may help you to improve your execution outcomes?
We also look at venues accessed and internalisation rates. I think there is a cost trade off between greater liquidity access and internal liquidity only, and hopefully this should be reflected in the cost we pay to the algo provider.
Many asset managers have already started employing FX execution algos. Do you expect to see more use of them being made in the corporate treasury environment and what can be done to increase their appeal amongst these important buyside firms?
I think they are a useful tool to add to a corporate treasury’s kit. They aren’t always better than traditional execution methods, and they have a cost to consider, but they certainly provide a further market access option which can be specifically designed to meet any trade requirements. The post trade output can also be easily reviewed by treasurers who may not have the market insights of a bank or a larger corporate dealing desk.
We also execute FX hedges for our commodity trading businesses which can be both significant in size and very volatile.
AI and machine learning are now being leveraged for algorithmic FX trading and associated analytics. How receptive are you to exploiting next generation technologies like these in your own treasury dealing operations?
We are always willing to look at developing technologies and market execution methods that will help us to achieve our goals. I think the use of machine learning and AI to decipher trading patterns, correlations and highlight optimal execution times is an exciting development.
In what ways are you likely to expand your use of FX execution algos still further in the future?
We have been increasing our use of NDF algos and hope to see more development of onshore algos in restricted currencies where liquidity can be more opaque than in majors. We are also looking at the development of FX Swap algos; as a large volume dealer of swaps at times, the ability to post passively is also an area we can see a cost saving.
We are always willing to look at developing technologies and market execution methods that will help us to achieve our goals.
