Skip to main content

Advanced Analytics in Trading Surveillance: Not Far Enough?

Recent cases of rogue trading and market abuse as well as new regulations are pushing financial institutions (FIs) to move from traditional supervisory models towards holistic trading surveillance models. Historically, FIs have responded reactively to new regulations with ad hoc controls and processes, but this approach created cumbersome inefficiencies and redundancies. FIs are now more aware of the importance of reducing the financial and reputational risks using a holistic surveillance model that analyses the trader behavior from different perspectives and blends all the information available, reducing the number of false positives and giving the supervisor the tools to understand the big picture.

Traditional analytics used by FIs to detect unauthorized trading are too simplistic as they scrutinize a limited number of scenarios (number of cancellations, limit breaches, P&L, VaR, trading outside “normal” hours,…) that do not adequately provide context for a transaction and generate too many alerts and false positives. The complexity of unstructured data does not help trading surveillance: in many firms, market and business data feed different systems and separate areas analyze the same trader’s activity independently as supervisors will look into trading activity while compliance teams examine electronic communications. Additionally, FIs are not able to use unstructured data from other sources like social media, websites or public reports.

Trading surveillance functions need to use all the information available, both internal and external, to find deviations from normal behavior. New technologies like Big Data solutions, Machine Learning functionalities, Natural language processing and Sentiment Analysis can be used to improve inefficiencies and gaps in trading surveillance, but the traditional analytics used in FIs are not enough to understand the trader’s behavior.  Traders involved in unauthorized trading are aware of the controls firms have embedded in their trading systems, mobile phones and chats but they can use other tools to avoid these controls. A small increase in trading activity with a specific counterparty may not trigger an alert, but a decrease in the number communications with that counterparty using official channels or other potentially suspicious activities – for example, the trader exiting the trading floor or the building right before booking the trade and not taking vacation or block leave – should make the trader’s supervisor suspicious.

Using holistic surveillance tools to mitigate unauthorized trading risks requires the use of advanced analytics that unveils patterns in trading behaviour to cover these violations. As we suggested in our last article on holistic surveillance, “a step-by-step approach towards holistic surveillance integration can deliver significant improvement to legacy surveillance activities”, and the first step must be revising and improving current analytics, giving context to supervisors to fully understand the trader’s behavior and detect hidden patterns of market abuse.“ The early detection of unauthorized trading is a priority for FIs, and while holistic surveillance is the ultimate goal, FIs must first attempt to better understand their traders’ activities before employing advanced analytics.