If things go wrong in financial services firms, those mistakes can end up costing firms very dearly. We’ve all seen the scale of the punishment: in relation to the FX trading concerns alone, firms have collectively paid over $10 billion in fines since 2013. According to the CCP Research Foundation, in the five years up to the end of 2015, the world’s 20 biggest banks incurred fines, legal bills and customer compensation of a huge £252 billion.
“Unauthorised trading incidents – such as those seen at Barings show us that balance sheet damage in the billions, or even complete collapse, can be triggered by one rogue trader,” says Rob Weston, Managing Director, Bank Risk Consulting at KPMG.
At the same time, there have also been renewed requirements from regulators over the past decade to address poor behaviour and market abuse.
The best way to mitigate these risks? Put in place effective trade surveillance systems. Not only do they keep the regulators happy and stave off punitive fines – they can also introduce greater efficiency and value-add into your business.
Banks have to monitor billions of communications, across multiple formats and locations, as well as hundreds of billions of trades. All this information then needs to be linked together to try and understand what has occurred.
It’s an expensive business, often made even more complex as banks are unable to approach the challenge head on. Some are already preoccupied with regulations such as MIFID II and the Market Abuse Directive/Market Abuse Regulations (MAD/MAR). Others are firefighting rather than adopting a rational approach: deploying urgent fixes to legacy systems as problems arise.
Many of these solutions not only come at a high price – they are not delivering the goods. Instead, they are building up a ‘technical debt’ that increases running costs and actually decreases a bank’s ability to detect risks. “Banks are facing a mish-mash of complex problems,” says Weston. “And they are not resolving the core issue”.
Traditional alert-based trade surveillance systems apply a set of rules to detect risks. Yet it’s not uncommon for more than 90 percent of alerts to be false positives.
In any case, finding a handful of relevant conversations in billions of communications is extremely challenging. The signal-to-noise ratio is made even worse by the number of false alerts.
These inefficiencies are rarely the result of any single issue, such as a defect in a process or a missing feature in a system. The problem is usually sheer complexity, made worse by time pressures, legacy systems, data volume and the low number of genuine incidents.
It doesn’t have to be like this. Automation and machine learning offer tremendous potential, not only to improve the way existing tasks are tackled, but also to meet much broader strategic requirements.
“Banks should be able to use their surveillance data for other analytics tasks,” says Lucas Ocelewicz, Director, Banking Risk Consulting at KPMG. “They make thousands of decisions when they review alerts, but don’t store and collate the high-value data those create. It could be a training set for machine learning, for example, so surveillance becomes more effective and cheaper”.
In other words, compliance is just the starting-point. “You add value by generating insight the business can use every day,” Ocelewicz says. “For example, surveillance data can drive better decisions on risk capital allocation; or highlight where traders achieve poor risk-adjusted returns."
Nor is it a case of starting over with a costly new system. “For instance, begin by adding some visualisation capability,” says Weston. “Tweak the existing technology for marginal gains. Don’t start with a leap to technology that you don’t fully understand, without the experience to challenge it.
“Eventually the technology life cycle limit will be reached and you can’t improve the legacy set-up further,” he concludes. “But by then you’re better informed on what to do next – that’s your quantum leap”.