Building an FX Algo Wheel: Head of FX Trading & Quant at Nordea Asset Management shares his story

“At Nordea Asset Management we look at FX automation as a tool to help us with three key business objectives – reducing operational risk, scalability, and freeing up trader time to focus on harder to execute trades. We have put a lot of effort into standardizing our use of algos and RFQ execution to support this. 

In order to reduce costs further we see significant value in automating order generation. So far, we are building this internally, but we use the core building blocks from our OMS and EMS providers to support us. If this is a priority for other firms then it’s critical that any OMS they use has the necessary functionally to assist them, or else they might find there are some bottlenecks that arise. When you have more manual workflow, you can get by with limited functionality in your OMS, but as automation grows, it becomes a more vital tool.

That being said, the biggest obstacle we’ve had so far isn’t related to technology. What we found most challenging was standardizing manual workflows in order to create common best practice which in turn translated into automated processes. Also, from a cultural perspective this is a game-changer for the firm. Portfolio managers and traders need to understand how they can work closer together and transform the order generation and execution process to make it more cost effective, and deliver better returns for our clients. 

What we found most challenging was standardizing manual workflows in order to create common best practice which in turn translated into automated processes.”

Establishing our algo wheel, which went live in January 2021, was a real milestone in this project. We see immense value in FX algo wheels, because they help us to effectively measure and compare the performance of various bank owned algos on an exploration and exploitation basis. Put simply, you can use the wheel to identify the best algos and then allocate more of your flow to them, but you can also balance this with constantly exploring new algos and testing them with smaller parts of your flow. 

This is incredibly difficult when you are using them on a discretionary basis. It also helps to scientifically set a foundation for when to use algos and in what currencies. 

There is inherent trader bias in manual algo selection, whereby if a trader decides a specific algo is best for a specific currency pair, then you will build up a large sample size with that algo for that pair, but not have large data sets on other algos to compare it to. The wheel removes this bias and allows us to build relevant sample sizes across different algos. We can then justify which ones we use in certain conditions. 

The wheel removes this bias and allows us to build relevant sample sizes across different algos. We can then justify which ones we use in certain conditions.” 

Another important part of being able to compare and benchmark algos is making sure they are standardised. When you customize an algo you introduce trader influence that is incredibly difficult to measure. We’ve standardised our algos into four categories (IS, VWAP, TWAP and Passive) and have standard parameters that we use for these across all our brokers.

As part of my role at Nordea I also sit as the head of a multi-asset quant team. We’ve been able to leverage some of the work that the team have done for our equities desk. One thing we learnt from this project is that the overall framework for FX and Equities are not too different. The main challenge is in equities you have many observations to collect, due to the fact algos are used for smaller trades in equities. However in FX we use them for larger amounts. This means we can collect less data and this makes the FX algo wheel much harder to establish, but the frameworks remain the same. 

We’ve standardised our algos into four categories (IS, VWAP, TWAP and Passive) and have standard parameters that we use for these across all our brokers.”

Smaller asset managers that have less FX flow than us will find this even more challenging. Longer term, if more firms want to adopt this model then data sharing through TCA providers such as the service that BestX offer will become very important. We’ve taken a lot of advice from them and took inspiration from a model that Pete Eggleston laid out for standardising algo usage when we undertook this project. 

We’re still in the early stages of using the algo wheel, and it will require consistent monitoring to make sure we are tracking the right metrics as the market and the algos we use evolve. The next step for us is monitoring the output from the wheel and using the findings to make decisions.”

We took inspiration from a model that Pete Eggleston laid out for standardising algo usage when we undertook this project.”