In the financial world, reputation is everything. Customers don’t just expect their data to be held safely and securely; they demand it.
Following an event of financial fraud or hacking, 30 percent of customers say they trust their bank less. 20 percent move bank while 51 percent say the bank should be held responsible for all losses.
A hack could mean a major loss to a financial institution, with customers potentially moving to another bank or expecting a pay-out.
With 389 billion digital financial transactions made every year, banks and financial institutions are turning to new technology to protect their customers. One approach gaining traction is the application of machine learning principles in fraud analysis.
22.8 billion reasons to change the status quo
In 2016, banks around the world lost $22.8 billion because of card fraud. That figure is set to grow to $32.96 billion by 2021. According to Marketwatch, “the majority of brokerages (88 percent) and financial advisers (74 percent) said they have experienced cyberattacks directly or through one or more of their vendors.”
Technology has made banking more accessible: we use our bank accounts daily to withdraw and transfer money, pay bills, and make online and contactless payments. But this access comes with a flipside: every transaction needs to be verified to ensure the exchange isn’t fraudulent.
Given the number of transactions they process, banks can’t manually enforce transaction verification. The traditional approach in the financial industry emphasised a mix of automation and human-level verification. Data was automatically analysed, fraudulent transactions were flagged and then checked in a monitoring centre.
The modern approach takes human error out of the equation and marries a bank’s datasets with the best analytical models. This more effectively prevents fraud.
Using machine learning to revolutionise bank security
Banks are increasingly using machine learning algorithms to analyse huge datasets and raise fraudulent issues. These algorithms aren’t entirely different from a predictive maintenance model.
Predictive maintenance spots anomalies in a process and sends a report to a nominated team of technicians. This data is built on with each incident and, in effect, the model becomes ‘smarter’ or more efficient.
With financial institutions, the algorithm spots anomalies in a similar manner. These can include instances such as large withdrawals or transfers to an account in a country a customer had no previous relationship with.
Essentially, the algorithm builds on the customer profile base to flag unusual patterns.
Not only is the algorithm better at identifying potential fraud incidents than the traditional model – it also provides far fewer false positives.
False positives are incidents where transactions are flagged as fraudulent when they aren’t. Reducing false positives increases the bank’s reputation, allows the customer to continue with their purchase, and reduces undue stress on the client.
Machine learning frees up resources
The true benefit of machine learning is in comparing the old with the new. We recently worked with a bank which utilised a system based on ‘rules’. If a transaction went over a certain threshold and was outside the usual patterns, it was flagged. The same incidents were being constantly flagged, but the system couldn’t learn.
Each flagged incident had to be checked by a member of the security operations team. Often, a cursory glance at the transaction was enough for the security staff to dismiss or action the alert. If a system is flagging thousands of false positives every day, it adds up to a huge cost in terms of resourcing.
After implementing the Statwolf solution to its fraud detection process, the bank reduced its losses because of fraud. It also had a reduced reliance on the monitoring centre as staff didn’t have to check as many false positives.
Above all else, it provided a more efficient process for the customer, helping to keep them happy and onside.
Are you interested in revolutionising your banking system with machine learning?
If you want to harness the power of machine learning in your business, Statwolf’s data science service can help, with advanced online data visualisation and analysis simply running in your web browser.
We offer a range of custom services to suit your needs: advanced data analysis and modelling, custom algorithm creation, and fraud analysis.
Want to make sense of your data? Download our comprehensive guide: The Predictive Maintenance Cookbook.
