The days when trading data and electronic communication were analysed separately are antiquated. The way forward, says Michael O’Brien, head of product management, risk and surveillance at Nasdaq, is to focus on monitoring the activity of traders and bringing in other sources of data that can add value to the alerts delivered to a firm’s compliance department.
“Just adding noise or more information doesn’t necessarily mean more insight. Sometimes it just brings in more noise and makes things less clear. It is crucial that we generate really good signals from an e-comms perspective and understand how traders communicate in chat channels,” says O’Brien.
He notes that using heritage technologies that detect certain words such as ‘fix’ can result in as much as 99% of false-positive alerts, which, rather than solving the problem, exacerbate it with additional irrelevant datasets.
But the use of artificial intelligence, machine learning and natural language processing tools can drastically cut down on false alerts by rummaging through the data to detect anomalies or changes in traders’ behaviour or risk profiles.
“As the application of these technologies in financial services has evolved over the last three to five years, we’ve seen customers reducing false positives by 60–70% using this more advanced e-comms monitoring technology,” O’Brien says.
In developing this holistic approach, Nasdaq partnered Digital Reasoning, a firm with its roots in the intelligence sector that specialises in electronic communication surveillance. The two firms, working in partnership, won Best e-Surveillance Solution Provider at the 2017 FX Week e-FX Awards.
“Within our holistic strategy, in phase one we were looking at bringing together two channels – trade data and e-comms – whereby we would take two existing alerts and bring them together in a meaningful way in context,” says O’Brien.
“In phase two, we’ve started to take these signals out of e-comms platforms and signals from trade platforms, to include trader profiling, communications profiling, know your client, gifts and entertainment, and taking all these signals and aggregating them on a per-trader basis,” he adds.
Until now, interest for surveillance tools mostly came from the sell side. But in the past 18 months, driven by regulatory pressure, an increasing number of buy-side firms such as hedge funds and asset managers have been looking into deploying such systems.
Using artificial intelligence to sift through the various channels of data to generate alerts decreases the likelihood of false positives and increases the probability of an alert containing actionable content. The ultimate aim is to establish whether there is malicious intent in a change of trading pattern or behaviour.
As the world of trading becomes increasingly electronic, surveillance challenges lie in old-fashioned, voice-based communications.
“Voice is the next frontier. The issue here is how do you effectively monitor voice communications to extract meaning, and identify collusive and manipulative language within voice,” O’Brien says.
The above article was first published on FX Week's website here.