Machine Learning needed on the trading floor
The list of Machine Learning (ML) use cases on a bank’s trading floor is long, and the financial upside of getting things right is phenomenal. The consensus on the Street is that a complete rebuild of sales & trading operations around data is no longer just an option but an urgent priority. As one investment bank CEO recently put it:
But rewiring the trading floor is a massive task and execution risk is high.
Among the plethora of firms offering solutions on the hot topic of data-driven strategies for financial institutions, those equipped with the right technology, quantitative skill set and sector expertise to drive measurable results are rare birds. In this post, we offer our take on the right approach to get things done.
Many opportunities, little time: new approach needed
The trading floor is a highly sophisticated, highly regulated environment where decisions made in real time have immediate economic consequences. Any innovation must be tested extensively before it can be brought to production.
Pressed by shareholders and the competition, banks are faced with a tight set of constraints:
- Many complex problems to solve
- A wide spectrum of complex AI/ML techniques to solve them
- Very little time to show measurable returns on investment
Solving this conundrum is possible but it requires a new partnership approach between banks and a new breed of independent fintech companies, powered by a dedicated data & analytics platform and led by financial market practitioners.
We recommend a business-driven, desk-by-desk approach to innovation on the trading floor. This approach requires the ability to test numerous algorithms on heterogeneous data sets and evaluate specific solutions through business-standard performance metrics.
Despite the consensus mentioned above, spearheading highly quantitative solutions with such a profound impact on functioning businesses is a big decision for senior management. To drive this change confidently, division heads must be equipped with a combination of:
- A business-friendly formulation of the AI/ML solution proposed
- A quantified impact measure using pre-agreed metrics
- The commitment to a technology transfer through the delivery of code or a turnkey expert system
- If required, a road map to production implementation
At qbridge, we draw on years of development of advanced AI/ML analytics for the financial sector to offer clients a wide choice of algorithmic options, implementable within months (not years) in the form of quick Proofs Of Concept (qPOC).
Our Analytics are provided as a Service (AaaS) to support every assignment, setting us apart from most generalist firms, as illustrated in the table below. Our (100% proprietary) analytics are based on open-source libraries and our data infrastructure is cloud-based, enabling platform-agnostic technology transfers. If needed, developments can be done on client premises.
We guarantee confidentiality on all aspects of our projects.
The 4 phases of a typical rewiring mission are:
Our market experts identify businesses with the highest AI/ML upgrade potential. This choice is validated by senior management.
This phase involves 3 steps:
- Definition of market-standard performance metrics (e.g. capital consumption, Return on Equity, balance sheet velocity, revenue per head, etc.). Metrics are validated by senior management and business heads.
- Design of a solution, including objective functions and constraints. This process includes a high-level specification of data requirements.
- Identification of the best-suited AI/ML algorithms for the job. This draws on an extensive repository of analytics and a dedicated data infrastructure enabling multiple test iterations.
At the heart of the value proposition is the delivery of an actionable solution through a transfer of technology (code, expert systems). This is achieved in 2 steps:
- Coding of a piece of kit to consume data and produce results.
- Delivery of a qPOC in the form of code or turnkey expert systems, accompanied with quantitative performance metrics.
A natural next step is for the client’s innovation lab, IT and business teams to convert the qPOC to a live production system. This more resource-intensive piece of work requires internal teams to work together, with the assistance of the fintech partner (if needed).