Deutsche Bank announced that they will launch an enhanced securities settlement service that identifies in-flight security transactions at risk of settlement delay, to debut in early 2021. The enhanced service will be using machine learning and anomaly detection features from SaaS company Elastic, and aims to avoid hefty financial penalties under the upcoming implementation of the new Central Securities Depositories Regulation (CSDR) in 2021.
Deutsche Bank’s settlement service will use Elastic’s machine learning (ML) technologies, so that the platform can move from real-time to forward-looking. The ML-enhanced service will detect the in-flight transactions that require actions and alert the bank’s teams before the transactions encounter issues.
“Our aim is to deliver a real shift in how markets view exception processing and to bring pre-trade performance to our post-trade operations. We can now detect transactions in real-time that previously would not be flagged as at risk, and divert our attention from the transactions that ostensibly appear to be at risk, but upon historical analysis have always matched in time to settle,” said Christopher Daniels, director of Data Products in the Securities Services division at Deutsche Bank, in a statement.
The analytics model underpinning the enhanced service is powered by Elastic’s anomaly detection feature, which considers seasonality, market variation and other changing dynamics to provide the bank’s operational teams with dashboards and action queues that are driven by a large set of factors that would be too broad and complex for a human to process.
“We have developed several dashboards covering liquidity, settlement performance, and risk and control, but the most recent innovations have been in running machine-learning algorithms in production to provide outlier detection. We’re using the platform to identify the most influential features that are more likely to cause a late or failed settlement, and to focus our data quality reviews on activity that does not synchronise with what we would typically expect from a particular cluster. It’s a very exciting time in our data roadmap,” said Daniels.