Trade Automation: A Mid-Sized Dealer’s Perspective
With Peter van Vught, Head of Dealing Desk, KBC Asset Management
Briefly describe KBC Asset Management, the profile of its dealing desk, and your role and responsibilities at the firm?
KBC Asset Management is the asset management entity of KBC Group NV. With EUR 200bn of assets under management, the firm holds a market leading position in several European countries, including Belgium, Czech Republic and Hungary.
The centralized dealing desk, based in Brussels, consist of 13 professionals responsible for multi-asset order execution, optimizing trading infrastructure, execution analytics and the management of the derivatives positions of its structured funds.
As head of the department, I’m primarily focused on the oversight, reporting and infrastructure and less on the daily trading activities. I aim to trade each instrument on a regular basis to keep up with the market and processes, but I leave the complex orders to the dealers.
What is the current state of automation on the buy-side trading desk (industry wide and/or at KBC specifically)?
For a medium-sized trading desk such as ours, a certain level of automation is needed to reduce operational risk, increase speed of execution, and improve post-trade execution reports and analytics, but a fully automated execution flow is not required and is not a goal as such.
With our recent transformation to a new order management system (OMS) and execution management system (EMS), we have been able to make very good progress in the area of automation, especially for equity order flow.
We have created an algorithm which allows us to automate 95% of our equity order flow. The algorithm is based on strict rules and uses several data inputs. First it calculates a liquidity identifier for each order using real-time market data and order instructions. Next it also evaluates live orders, fee schedules, restrictions, etc. to determine if the order needs to be executed via Electronic / Program / Cash trading. In case of electronic trading the algorithm will also select the optimal algo strategy. Finally the algoritm will select the preferred broker based on historical TCA results. The time saved by this higher level of automation allows the dealer to focus on the 5% of orders which need extra attention.
However, we choose not to automate the entire equity execution process. The dealer still needs to release the order to the pre-populated destination. We see limited added value of fully automating this process. As this can be done in 1-click for a bulk of orders, the time that can be saved is limited, while in the current process the dealer is still fully in control of the order flow.
The integration of platforms like Liquidnet and Tradeweb to our Equity/ETF Execution Management System has helped us to use different trading protocols in a semi-automated way. Also for these functionalities, the execution is not fully automated and requires one or two clicks from the dealer to execute the order.
Which processes/workflows on the trading desk still need to be automated? Which may never be automated?
Where we still see room for improvement in the equity flow is the integration of the Indication-of-Interest (IOI) in the automated workflow. When a broker or liquidity-seeking algo is working on a high-touch order, it would be very beneficial for the dealer to be automatically informed if another broker has issued a new IOI on the stock. Currently this is still a manual process, but discussions are ongoing with our EMS provider to have this process improved and better integrated.
This will only work if good quality, real-time and tradable IOIs are integrated in the EMS. For smaller asset managers, this creates an additional challenge as brokers might be more reluctant to adjust their processes/IOIs for smaller clients.
What is the state of trading-desk automation in other, non-equity asset classes?
Within fixed income, the corporate bond market remains an area where we see potential for further automation. Even though we’ve made a start with rule-based execution on small-size corporate bond orders, for the majority of corporate bond orders the dealer still needs to manually select counterparties for the RFQ on the platforms or contact counterparties in their search for liquidity.
We are currently working on the integration of axes in our Order and Portfolio Management System. At the desk we used to have a proprietary application that aggregated all axes which we directly received from counterparties and other providers. This is now integrated in the OMS which makes it also easily accessible to the fund manager who can use it in their portfolio rebalancing.
To save critical time and to make full use of the available execution and market data, we’ve initiated a project with the Data Science team of KBC Group to develop an algorithm which would select the counterparties with the highest likelihood to win the trade in the RFQ. This would support the dealer in its counterparty selection and save time during busy trading moments.
After several months in production, the algorithm is currently being evaluated. During the evaluation we’ll especially focus on two main aspects:
i. In volatile markets, algorithms based partly on historical executions are generally less accurate. How does the algorithm hold in these situations?
ii. The model currently runs on a stand-alone Python application and the dealer still needs to manually take over the outcome on the trading platforms. How can we fully integrate the self-developed model in the current OMS or on trading platforms?
With recent developments in corporate bond markets, we have no doubt the level of automation will further increase. Whether we’ll further elaborate on our own models and tools or rely on vendor-based solutions like the rule-based execution on the MTF, needs to be seen.
What are KBC’s challenges with regard to automation, as a mid-sized dealing desk with less scale compared with larger rivals?
We can indeed be considered as a mid-sized dealing desk, and we don’t have the scale of some larger global players. But this also creates some advantages.
It gives the dealers the opportunity to learn from different asset classes. It eliminates a silo mentality, encourages cross-learning and improves collaboration. And this clearly benefits everyone. New perspectives help drive our business forward.
Being part of KBC Group, which strongly focuses on innovation and digitalization, also gives us access to experts in those dedicated fields. As with all firms, resources are scarce, and a good business case needs to be available. But if we have a case, we have the resources.
With the increased level of automation, we see changes in the tasks and required skills of the dealers. System and process knowledge, data analytics, collaboration (with IT or vendors), etc. will all become more relevant if the level of automation increases. This is where we see some challenges as we can’t just have several colleagues in the team dedicated to data analytics or process improvements instead of order execution. We expect from the dealers that they take up some of these additional roles and continuously develop their skills and competencies while maintaining their primary focus on trading. This is a challenge for the dealers but being able to use the daily execution and market insights in execution analytics, process evaluations, IT discussions etc. definitely has merit.