The digital revolution of the last thirty years has transformed almost every aspect of the financial services industry. But amid this sea change, one constant has remained: the fundamental relationship we have with the technology itself. Simply put, computers have always needed people to tell them what to do.
Today, even this situation is changing. By applying complex algorithmic modelling, computers have for some time been able to explore and learn from data faster and more comprehensively than humans can. This means they can anticipate more accurately future outcomes and, as a result, make better operational decisions. Unlike its human counterpart, a machine-based ‘analyst’ has a potentially limitless capacity for processing and analysing information. Crucially, it can also generate real-time results without applying any form of typical human bias. This brave new world of ‘machine learning’ has the potential to help a data-laden bank or financial institution seriously up its game.
For years, banking analysts have been using statistical models to understand customer behavior, albeit with a less flexible and more expensive older generation toolset. Tomorrow’s gains are more likely to come from pushing the analytics workload to ‘the machines’ that handle more and more predictive and prescriptive use cases. Resultant forecasts can inform all manner of internal adjustments, from when it will be necessary to replace failing equipment, to identifying which employees will be most productive as they mature in their roles. Huge value also lies in how products and services can be enhanced by machine generated insights that help identify which transactions are most likely to be fraudulent, gauge a customer’s susceptibility to default, or even their propensity to take their business elsewhere.
But the value is not only in the ‘what’, it is also in the ‘when’. Machine learning can identify and create advanced warning systems that enable early detection of delinquency, credit card fraud and customer attrition, helping the bank to reduce risk. In a complementary bottom up approach, the same machines can be let loose to explore data and reveal previously unknown relationships that can pave the way to novel products and services.
But for all its potential, commercial adoption of machine learning in financial services has been slow. Data science teams, with academic backgrounds, have struggled to adapt to the commercial world of financial services, due to issues like rigid legacy IT platforms and the sector’s overall risk-averse working culture.
On a commercial scale, machine learning also requires investment in massive processing power. The cloud offers a viable solution here, but until access gateways can be completely secured and regulatory ‘safe harbor’ issues resolved, many banks are understandably insistent that their data remains safely behind their firewalls.
Fortunately, the machine learning industry is responding to overcome these barriers. A new breed of bank-friendly startups are emerging, whose cloud born platforms overlay the discipline’s algorithmic wizardry with on-premise (or hybrid) deployment options, clean and functional APIs, and a simple, intuitive user-interface that enables even non-experts to get up and running quickly.
By way of illustration, Poul Peterson, CIO of machine learning specialist, BigML, has recently used the company`s machine learning platform to set up and conduct a live analysis of public data from peer-to-peer loan company The Lending Club. After a few minutes of setup, the platform analysed the company`s public loan book information and immediately generated a risk-based model through which the repayment behavior of existing and future borrowers could be predicted. These predictions can give the lenders a new level of risk-based insight before they part with their money.
The Holy Grail for banks lies in deploying ‘pure’ machine learning, where the machines are unleashed on the data without restraining parameters. This is when the field’s true potential can be realised. For now, however, the commercial gains for banks and financial institutions lie in the ability to upgrade existing processes and help inform current decision making.
As adoption increases, machine learning is likely to disrupt and reshape its host businesses more fundamentally. As the user friendly and massively scalable next generation analytics platforms emerge, it has the opportunity to do just that.
About Amir Tabakovic
Amir Tabakovic has spent the past twelve years pushing the boundaries of digital financial services and technology. Amir currently serves as VP, Business Development at the machine learning pioneer BigML, where he is responsible for driving business growth and identifying development requirements for the firm’s machine learning platform. Amir previously served as Head of Market Development at PostFinance, the fourth largest retail bank in Switzerland where he established the bank’s mobile engagement strategy.
He is an honorary lifetime member and former member of the board at global mobile financial services industry association, Mobey Forum. Here, Amir chaired the Mobile Wallet Workgroup, which published numerous papers charting the evolution of the industry and providing recommendations for banks seeking to establish a position in the market. He now chairs the association’s Predictive Analytics Workgroup.
About BigML
BigML is built on a passion for creating systems that learn from data. Its goal is to make machine learning simple and beautiful, enabling companies to uncover the hidden predictive power of their data. The firm’s international team harnesses large-scale machine learning, distributed systems and data visualization and delivers them to businesses via a service platform that is elegantly simple, easy to use and massively scalable.
In 2015, BigML was named a data science Cool Vendor by global analyst firm, Gartner.
Founded in 2011, BigML is headquartered in Corvallis, Oregon.
Every day we send out a free e-mail with the most important headlines of the last 24 hours.
Subscribe now