Previous proposalCase study: Outbound logistics optimization for multi depot problem with time window
Next proposalMachine Learning Model Management with MLflow
Data Science for the discretionary managers: Lessons from a 60 trillion$ traditional industry resistant to change and facing the quant threat
Investment management is a 60 Trillion$ industry, and despite the recent advancements in data science and machine learning, still remains fairly discretionary. Untill recently, less 20% of the funds called themselves quantitative.
However, there is an absolutely massive transformation taking place right now within the discretionary investment management industry. Quantitative and systematic strategies have produced far more consistent returns over the last few years and investment assets are flowing out of discretionary funds at an alarming rate.
Discretionary managers have finally woken up, and are now scrambling to understand what’s taking place and how they must change in relation to it. Many will not survive the shift. Others, who move quickly and efficiently towards building quantitative processes will take advantage and be better off for it. The key to make success will lie in building the right infrastructure, hiring for the correct roles and have cross functional support.
We will dive into the following themes and how institutional managers can begin to effectively redirect themselves:
- Investors are finally aware of asymmetric risk they were taking on with active discretionary mutual funds and hedge funds.
- Most classic systematic strategies based on price; volume and fundamentals have been arbitraged out and there is now an arms race to build new strategies with new data sets.
- Discretionary managers are scrambling to incorporate new data sets, but lack the understanding of how to analyze their efficacy and more importantly, how to incorporate them into their discretionary trading processes.
- The organizational structure of discretionary management teams along with the type of people they hire is broken and outdated for today’s challenges.
- Companies cannot just hire a bunch of data scientists, tell them to work with 50-year-old fund managers with MBAs and hope that magic will ensue. The two sides simple do not speak the same language
- Fund managers aren’t educated as to how all of this works and often distrust the data and signals coming out of the process
- Building the right infrastructure will remain pertinent to surviving this shift
- Firms should invest in a centralized infrastructure capable of acquiring new data sets, doing basic descriptive work and making it available and reliable enough to have PMs use across the firms.
- In addition, firms need to build cross functional teams on the PM’s desk, customized to their style and support them with the data and infrastructure team at the top. Fighting over centralized quantitative research capacity with other PMs will lead to a disaster.
- Explainability is the key! Firms need to ensure there is a role/process in place that has a deep understanding of the PMs process so that they can work in coordination and also a strong understanding of quant processes to bridge the gap.
- Finally, PMs need to upskill themselves with a basic understanding, in order to effectively communicate and run their teams. If they aren’t educated as to how all of this works, they are never going to trust the signals coming out of the process when the time comes to make buy and sell decisions?
I am the CEO/founder of Auquan. I have 8+ years of global experience in finance with Deutsche Bank in Mumbai/New York and as a derivatives trader with Optiver in Chicago, where I was the first (and only) female trader in the company! At Optiver, I traded volatility arbitrage strategies and was involved first hand in making the shift from discretionary to automated trading. Since 2017, I have been employing new and cutting edge ML and Deep Learning techniques at Auquan to solve financial prediction problems for hedge funds and asset managers.