Video: The Future of Trade Surveillance



In this third video in our series on trade surveillance processes and technologies, we take a look at what the future holds, in particular how artificial intelligence, machine learning and behavioural modelling approaches can help predict and prevent manipulative or abuse market activity.


William Garner – Charles Russell Speechlys
Taras Chaban – Sybenetix
Nick Gordon – Certeco
Michael O’Brien – Nasdaq

Hosted by The Realization Group

For more in-depth discussion read our Financial Markets Insights article

Mike: Hello and welcome to Financial Markets Insights. Increasing regulation around financial markets, particularly since the introduction of the market abuse regulation and directive, has led to participants having to take a smarter, more proactive approach to spotting abusive trading patterns and technology is playing an increasingly important role.

In this video, the third in our series on trade surveillance, we look at how firms are exploring more sophisticated use of technology to pre-empt abusive or disruptive trading behaviour. First of all, what are the regulators actually looking for?

William Garner: With the development of markets and electronic markets, the expectation is that you do not have manual checks so that every firm, wherever they sit in the chain, if they’re sell-side, buy-side, high frequency trading firm, market maker, that they have some form of electronic surveillance. It would certainly be unusual not to.

In terms of what that surveillance is and what it does, it is, obviously, not prescriptive but what is essentially expected is that it is fit for purpose, however you go about ensuring that’s the case. The expectation of the regulator is dictated by the nature and scale and size of the firm. The more sophisticated the firm, the more their impact or likely impact on the market, the more the expectation is, on the part of the regulator, that they would have more sophisticated technology in place.

Mike: As firms start to look at more sophisticated technology for trade surveillance, what does that mean in practice?

Nick Gordon: The key point to get across is to make sure that any technology you take on board is fit for purpose and provides you with the benefit that you need as a compliance team in order to work effectively within your organisation. Too many times I’ve come across companies that are promised the earth through technology when, actually, a very careful evaluation needs to be undertaken in order to make sure that it meets your requirements.

It just boils down into the same issue that every compliance officer has around surveillance, which is how many false positives and how many false falses am I getting going through my systems? How can I improve that? One of the big areas that we’re constantly being asked to explore is the area of big data where you can actually take in all your trades and do the analysis that some of the big data providers say that they can do around surveillance.

Our perception is that this is actually confusing the market place because there’s only a certain amount that you can do with big data and, as we move to more instant management, big data isn’t the solution.

Mike: If big data in and of itself isn’t the solution to this, does the future of trade surveillance lie more around the use of Artificial Intelligence and Machine Learning? If so, how will they potentially work with smaller data sets?

Taras Chaban: If you look at actual techniques for Artificial Intelligence and Machine Learning, they haven’t really developed hugely compared to some of the work that was being done even 10 or 20 years ago. It’s more about availability of data sets, about computer power but the most important work in any data analysis, be it Machine Learning or AI, comes, actually, at the beginning when we figure out what features are important, what data points are important and then apply the techniques to them.

The majority of people look at how can you load lots of data, throw a lot of computers on it and let them figure out what’s going on. We look at it from a behavioural science perspective so we say, “Actually, before you look at the data and before you throw it into the computer, it’s important to understand the job, it’s important to understand how people do their job and why they make certain decisions”.

Only with that, with observation, observing what’s going on, how people operate, what constraints they have, you can actually understand this data better and, therefore, arrive at better analysis and better conclusions.

Mike: Behavioural analysis is likely to be a key aspect of trade surveillance in the future so how can it be made more powerful and more accurate?

Michael O’Brien: I think the idea of utilising Machine Learning techniques, what that definitely lends itself to is this idea of a community-based approach to identifying anomalies and learning through time what is anomalous and what does indicate unusual risk and what doesn’t.

We have a baseline level where we benchmark using Machine Learning and Artificial Intelligence techniques to cluster our customers and create a profile across our customer base of this particular trading profile. We bucket those customers together and, through that community, we can share anonymously and abstracted how that particular group of customers have their parameters set across different alert scenarios.

That’s a very small step, which our customers find very useful because then they can look at where they sit relative to their peers in terms of how they’re configuring their alerts and what they have enabled. We’re now starting to talk to our customers about how could we, effectively, do that learning across the totality of our customer and analyst base, which extends to thousands of analysts across multiple markets?

They’re certainly very open to that idea and interested to pursue it. There are technical challenges about how you abstract that and anonymise it and so on but it’s definitely where everyone is thinking.

Mike: What is the future of trade surveillance? Automated rules-based systems do serve the industry well but as they become augmented with more sophisticated AI-based systems that use Machine Learning and behavioural modelling to identify and pre-empt abuse then, hopefully, the markets should become safer for all. Thanks for watching. Goodbye.