Few things in risk control are more pressing than being able to detect rogue trading. Banks have increasingly paid fines over recent years for bad actors that have slipped through their radars. And when one bank gets hit, the share prices of all tend to slip as well.
But detecting unauthorised trading often means sifting through a sea of data. That means applying constantly improving technologies like AI and ‘the cloud’, but also preparing the right framework at the start.
Panellists at 1LoD’s “Unauthorised Trading and ‘Joining the Dots’ Across Controls” discussion emphasised the need to focus on the most meaningful metrics first.
Rogue traders will invariably find ways to walk around the systems banks have put in place, said Ian Dalglish, managing director and head of global markets, prudential, at ICBC Standard Bank. “We use systems and we want to use systems, but it’s also the quality of the data and the amount of data you have that dictates the effectiveness of the tools,” he said.
So starting simply is often the best approach, before leaping to technology as a silver bullet for detecting rogue traders. “Start with the data you’ve got,” said Chris Palmer, managing director, CIB controls, at JP Morgan Chase. “Cleanse that, get it as accurate as you can. Most importantly, work with the front office” to ensure you’re focussing on the right risks and controls. “If you jump too deep into technology to start, you get complaints from the users saying they want further functionality and you end up trying to run before you can walk.”
That can result in a flood of data that, in the long term, may only hinder the process. “Everyone is struggling under the weight of flags and alerts, because the scope of regulations and asset classes is expanding all the time,” said Michael O’Brien, head of product management, risk and surveillance business at Nasdaq. “You have to focus on efficiency first,” he continued. “How do you apply these technologies to guide the analysts and teams on where to focus their attention?”
Sheer brute computing power can have big benefits, here, said Gordon McKenzie, vice president, EMEA and APAC for Digital Reasoning. AI systems can learn under supervision and be more effective at identifying intent than classical systems, he said. “They can clear up a huge amount of the noise the classical systems would generate,” said McKenzie, adding that Digital Reasoning has been able to reduce up to 90% of false positives for clients as its system’s learning abilities improve. The firm has also increased true positives by up to five times, he said.
But it’s important to remember that conduct starts with knowing your business, said Dalglish. “Nothing is a substitute for the desk head knowing what their traders are doing. Generally, we’ve found that with rogue traders, if there are issues with their trading, there are issues with other parts of their day-to-day.”
Discover more content by 1LoD
90% of Global Banks Expect the Front Office Controls Mandate to Expand
In Conversation: The Reality of Conduct Risk
The FCA’s 5 Conduct Questions
Good Conduct is Good for Business