Trader as Technologist: Buy-side Interview

J.P. Morgan Asset Management’s Head of Trading, Kristian West, shares the role technology plays on the trading desk, including the time given to technology, evaluating tools and conveying that value to portfolio managers.
IT and the Trader’s Diary
The technology function in many firms was previously seen as a necessary evil, whereas now it is seen as part of the core function of what we do. At J.P. Morgan Asset Management, we have eight technology specialists on the trading desk, so quite a lot of time is spent working with IT by virtue of them being on the desk. The trading function has become much more technology intensive and we wish it to be at the forefront of what we do. As a percentage I probably spend around 30% of my time specifically on IT related initiatives. We spend a significant amount of time thinking about and planning technology initiatives for their effectiveness and value as well as best practices.
In addition to this, there are a variety of obligations that keep me away from the desk, such as regulator meetings, compliance meetings, customer presentations, broker meetings, client reviews, TCA reviews, etc. We also have to ensure we have the relevant oversight and controls over our trading practices, which requires us to focus on relevant due diligence measures. Due to MiFID and various other potential regulatory changes, we spend a lot of time making sure that we have oversight and knowledge of all available options. We then take a view of what we think will be the likely outcome and focus our attention accordingly. Again, this requires the business and technology to work very closely together.
Communication with Portfolio Managers
The central trading function is responsible for achieving the best possible outcome for the customer. There is no specific guidance from investors on how to trade or where to trade. However, from a communication perspective, there is a great deal of dialogue when we have an order on the desk. Certainly, if it is a multi-day order, then as much communication as possible is encouraged so that there is a fluid relationship between the investor and the trader, maximising our trading opportunities. More formally, we have monthly and quarterly reviews with all the CIOs and investment teams to discuss market activities, flows, broker relationships, transaction costs, etc.
We are able to optimise our performance when we understand the intentions and motives behind the orders. This allows us to adjust our level of participation in the market. In addition, we need to consider upcoming events or corporate actions, as acting quickly helps reduce market impact and slippage.
Active or Passive? When to Pick up an Order
We have a defined flow process which is focused on liquidity, so if we have flow that meets certain criteria, it is automated and no trader is involved with it. The characteristics of the strategy that is chosen can actually be defined by the investor, allowing them to define how aggressive they want to be. With regards to allocation of flow, the OMS knows which area of the trading desk to send orders to and the automated engine will take orders that it feels it can execute. The automated engine then takes over and will alter its behaviour over the life of the order, depending on what is happening in the market.
A large chunk of business is fully automated, but if it does not fit defined criteria, it will go to the program trading team who will try and use their liquidity and the liquidity exposed to them to minimise market impact. If the order is too large (we have certain thresholds) or there is no natural liquidity in the market, then it will go to the single stock trading team. At that point we speak to specialists in those names. As a result of this process we have changed the way we interact with the market. We take on a lot more ownership and responsibility for the quality of execution and thus, less is left to the discretion of the broker. So whilst our execution process is not fully automated, the allocation process almost always is.
Value from Brokers
As mentioned the way we interact with the market has changed significantly. From an electronic perspective, value is increasingly found in customization, so how do you change the algo to better suit the individual client’s needs? We see this customization as an advantage, because wherever we automate our flow, we can auto-select from these customized strategies. From a broker’s perspective, there is a lot more customization happening so as to differentiate one from another. There is a fairly standard selection of algos that is consistent across providers, and it is these that can be best tailored to suit specific investment needs. At that point one can compare and analyse the performance of the customisation. FIXatdl is used for all our automated strategies, and we are moving our other strategies on to that methodology.
Benchmarking
When the time comes to evaluating our algo trading and strategy selection, we look at multiple characteristics. The base benchmarks we use are slippage vs. IS, slippage vs. VWAP, or Participation Weighted Price (PWP). More granular metrics (how often we cross the spread, how often we fill at midpoint, the average execution size, the frequency of execution versus our peer group, where we have executed) are used to evaluate how the strategy is behaving, how much market access it has and which strategies are generally working better. There is a great deal of analysis conducted on venue and strategy performance.
This is an in-depth process where we look at every child order’s execution profile to see where, on average, one strategy works better for us than another. It could be that one strategy is tuned to participate a lot more in internal market making platforms, in which case it could be that it is crossing the spread on very small executions frequently. Alternatively, the average execution size on another could be multiple times its competitor, thereby capturing much more of the spread on lower trade frequency. Here, it could give us a better execution performance depending on the characteristics of the stock. There are varying levels of metrics we look for from an execution perspective, and it can get rather complex.
Thankfully, we have a full time quant/TCA execution researcher on the desk. His background is in technology and role it is to analyze the data and constantly scrutinize and tweak the usage of different tools. As part of that process, we use TCA to ensure the execution profile works for either the investment product or the investment individual, so that our execution style is tuned to what they are trying to achieve. Benchmarking is not just about comparing how we did versus the market, but whether the algos are tuned to the investors’ intentions and goals for a particular order on a particular product. The process necessitates being heavily involved in the investment process as well as the trading process. As a result the overlap between investment management and trading is increasing.
Internal Innovation
Through work we have done around automation and transaction analysis, we have an increased focus on innovation. We have an innovation council within J.P. Morgan Asset Management that encourages the development of new technologies and thought processes. Over the last few years, we have added to the quality and quantity of technology staff within our operation, and this can be seen when visiting the trading desk. We work very closely with them to show them the fruits of their work, so they better understand what it is we are trying to achieve.
Anecdotally, we have recruited from the internal technology department to work within the trading team, which allows us to do more frontend customization with our EMS and algos on an ad hoc basis, rather than defining it as a separate project. It is quicker and more efficient to take someone from IT and put them on the trading function, rather than run through typical project management processes involving additional resources. Assuming it does not have a material impact for the back end, completing those projects in real-time on the trading desk means we can turn things around a lot quicker.