Machine learning: The next frontier for financial risk management?

The financial services industry has quickly embraced the power of data analytics. The ability to leverage crucial insights from huge datasets enables businesses to predict consumer behaviour, set strategic goals, identify and manage risks, and much more.

 

Machine learning is typically described as a method of data analysis, but with a number of key features. Essentially, machine learning involves computers that don’t require specific programming to perform certain tasks because they are capable of learning through experience via complex algorithms.

 

Google’s development of driverless cars shows machine learning in action. Programming a computer to account for the infinite real-world scenarios that occur when driving is impossible, so autonomous vehicles accumulate driving skills using data analysis and ongoing test drives. Until recently, Google’s cars had never been responsible for an accident in which they were involved, showing the potential of machine learning.

 

The search engine also recently created a computer that beat the world’s best Go player. The Chinese board game is often considered the gold standard for testing artificial intelligence due to the number of potential moves and strategies involved.

 

These are some of the current ways that innovative companies are pushing the boundaries of data analytics. But how are financial services firms – and specifically their risk managers – able to use machine learning?

Machine learning and financial risk management

A 2015 report from McKinsey & Company claimed a dozen European banks had made the shift from traditional statistical analysis modelling to machine learning. The results have been impressive, with some organisations increasing new product sales by ten per cent, while churn and capital expenditure have both declined by 20 per cent.

 

Common applications include microtargeted models that can predict the most likely consumers to cancel services or default on credit. These early warning capabilities ensure banks can intervene before problems escalate. Importantly, these systems can both predict behaviour and learn to understand how people react in certain situations.

 

Meanwhile, a new McKinsey study highlighted a number of data trends that are expected to arise in the risk management industry. As the regulatory burden broadens and consumer expectations rise, organisations will need ways of automating cumbersome processes to streamline the customer experience and manage risk without human input.

 

“Banks are experimenting with self-learning algorithms in credit underwriting, monitoring, and credit-card fraud detection, with encouraging results,” the report stated. “Advances in behavioral economics will also help risk managers make better choices as they learn to recognise and eliminate common biases from their decisions.”

Preparing for the future

Machine learning is already gaining momentum worldwide. Gartner identified it as a top ten strategic technology trend in 2016, with advances occurring rapidly. The analyst firm said businesses must find ways to actively leverage machine learning if they wish to gain a competitive advantage.

 

According to McKinsey, financial organisations must digitise their core processes between now and 2025. By this date, most companies should have minimised their manual interventions, with modelling, automation and standardisation becoming firmly entrenched. Risk managers should therefore look to take a proactive lead in suggesting the introduction of new technologies.

 

Credit applications and underwriting are the key areas where machine learning, and data analytics in general, will have an initial impact. The outcomes will include cost reductions, increased efficiency and less onerous customer experiences.

 

Fraud is also an important growth area. IBM Research has developed a modelling technique that utilised machine learning to analyse historical and real-time transaction data. The system enabled one US bank to improve fraud detection by 15 per cent, as well as reduce false alarms by 50 per cent and boost savings 60 per cent.

 

“Risk functions should experiment more with analytics, and particularly machine learning, to enhance the accuracy of their predictive models,” McKinsey’s report advised. “Some financial institutions have already achieved significant model improvements, leading to better credit-risk decisions.”

Making the right talent choices

The skills and infrastructure required to facilitate new data techniques can be considerable. Risk managers will not only need data analytics knowledge and abilities, but also the interpersonal skills to drive the necessary changes across various departments that handle sensitive and proprietary information.

 

Furthermore, McKinsey argued that the personal touch will always be needed, regardless of the level of automation. This is because organisations must ensure that technology is applied ethically and appropriately.

 

Machine learning may still be in the early stages of innovation, but the rapid evolution of technology means businesses should prepare for digital disruption now. The risk managers that possess the right data infrastructure sophistication and experience to harness machine learning benefits can achieve dramatic performance improvements.

 

Our Market Reports combine our review of the prevailing conditions in the risk management recruitment market together with the results of our latest employer survey.

 

Image: Olivier Le Moal via iStock