There is a renewed interest in machine learning techniques that makes it possible for computers to extract valuable insights from large data sets without being explicitly programmed. The availability of cheaper, more powerful computers and the decreasing cost of data storage have meant that machine learning algorithms can now work with vast quantities of complex data and deliver high value predictions in real-time without human intervention. In the last decade, machine learning has been applied in many areas such as web search, recommender systems, credit scoring, fraud detection, stock trading and many other applications.

Predictive analytics is concerned with the prediction of future trends and outcomes using approaches such as machine learning and statistical methods. Machine learning techniques have become increasingly popular as a means of generating predictive analytics due to their ability to identify the factors that lead to certain outcomes from within large data sets.

Leading companies in the financial services industry are already using machine learning and predictive analytics to gain new insights into customer behaviour. As a result, they can deepen existing relationships with their customers and build new ones, both leading to increased revenue opportunities. This ranges from the ability to retain clients by identifying those that are at risk of defecting, to providing a more tailored service. Machine learning and predictive analytics are set to become powerful tools for banks’ sales and trading operations.

Surprisingly, this could be accelerated by the growth in financial regulations that are requiring banks to improve their data management capabilities. The data warehouses that are built to support these needs have the potential to provide far more value than simply what is needed to continue in business.  As well as allowing monitoring and analysis of internal behaviour, they will also serve as the platform against which these algorithms will run to deliver customer analytics.

This data will also help banks to take a more client centric approach. Traditionally, banks have organized themselves as silos with limited information shared between their businesses. Machines learning techniques and predictive analytics will allow them to gain deeper insight into the customers across all the products they offer. There is however, a fine line in how this information should be used. Sophisticated clients will become wary of exposing too much of their business to a single dealer to avoid the risk that they can “reverse engineer” their trading strategies. It also raises the question of whether targeted pricing decisions based on behavioural patterns mined from the data could be interpreted as front running the client. Therefore, some clarity is required in where the boundary lies between serving the customers needs and taking advantage of them.

Whilst machine learning techniques can be applied to a myriad of problems, it is client behaviour analysis that has the greatest strategic potential. Unleashing machine learning algorithms on the wealth of data points that can be captured around client transactions through both electronic and voice channels will begin to unearth patterns that will enable banks to shape their product offering and organisation to best support clients.  Therefore, banks will be able to use their workforce more efficiently through a sales force that pursues more targeted opportunities and trading desks that prices optimally.

Clearly, this will be one of the major battlegrounds as banks seek to exploit their data. Those that lag will be at a serious competitive disadvantage, missing out on revenue opportunities as well as the cost efficiency that will come from maximising their resources. There is already a wealth of tools available to firms of all sizes, levelling the playing field, ranging from big data platforms such as Hadoop and Spark, open-source machine learning libraries like Scikit-learn [1] for Python or Weka [2] for Java, and cloud platforms tailored to machine learning such as Microsoft’s Azure [3].

It is not surprising that a British Government report into establishing the UK as a world leader in financial technology identified machine learning as one of the key disciplines [4]. Central banks and industry regulators are also recognising the opportunity that exists in the data that they now collect [5]^,[6]. The large financial institutions need to avoid being left behind.

Whilst there are companies that are advanced in their use of machine learning and predictive analytics this is still in its infancy within financial institutions that have traditionally been heavily relationship driven. As a result, there are significant advantages open to the early movers who build up the capability to collect, manage, and utilize their data. This will create significant demand on the skills required to support this, which is already in short supply today [7]. In particular, these will include a combination of domain expertise, technical knowledge and analytical skills. As highlighted in the recent best-selling book  “Big Data” [8], machine learning has the ability to expose the what, but will not give the why. This will still be left to the expert interpreter for the time being. As a consequence of this, many firms will turn to products that deliver this capability quickly and present the information to their business users in a clear and consumable form.