The banking industry has progressed significantly in the technology arena and has made some big changes to propel itself into the digital era. And it’s not just managing money and simple online banking where technological advances come into play. Spending money has become even easier with online payments and contactless payment systems, whether using the bank-issued card itself or a mobile device equipped with near field communication (NFC).
All of these developments mean convenience for the customer, but for the industry, it exposes new risks with more entry points and opportunities for fraudsters to attack. In early November, Tesco Bank reported £2.5 million in funds being taken from 9,000 account holders.
Details of precisely how the scheme was executed are still unclear, but regardless of the fraudsters’ exact methodology, the important question is why was the attack successful? Given the scale, how was the problem not detected more quickly, and how can other banks do better?
The Age of Algorithms
Despite the banking industry’s advances in customer-facing technological offerings, many traditional banks have unfortunately fallen short of embracing the latest technology when it comes to internal processes and infrastructure. This is in stark contrast to digital powerhouses like Google, Apple, or even smaller startups who are entering the banking space in the full context of the online and mobile world into which they were born.
In particular, some traditional banks have not yet embraced data science fully in a way that allows them to compete in the so-called age of algorithms brought on by the likes of Google and Apple. Data science is the collaborative discipline - the combination of people, data, tools, and processes - used to transform big data into actionable insights and business innovation.
In other words, data science can be the difference between a public relations nightmare associated with a Tesco-scale breach and a risk quickly identified but stopped before affecting customers. In the banking industry, fraud will happen (there’s no eliminating fraud risk when money is at stake), so the question becomes how to effectively and efficiently identify and minimise it, and there’s still much more work to be done in this arena.
Data science is the answer
Financial institutions have the unique advantage of deep, historical troves of customer data; we’re talking massive stores of day-to-day transactional customer information. But for many banks without a developed data science strategy, this asset goes untapped. In some cases, this may be due to data sensitivity and regulatory requirements. Banking is undoubtedly one of the more difficult industries in which to operate with a data-driven approach.
But even so, it’s something that can (and must) be overcome. Leveraging data mining and predictive modelling is literally sink-or-swim for detecting fraud in real time, as we see in the Tesco Bank case, and it doesn’t stop there. Banks and financial institutions using data science are able to stay ahead of competition and:
- Personalise offers by using machine learning to continually identify customers likely to actually buy or use the service being offered.
- Minimise operating expenses by automating tasks otherwise done by humans or bringing efficiency to paper-intensive workflows.
- Expand markets by using big data to accurately assess risk and serve those markets that are traditionally underserved.
- Create disruptive new products by using big data to understand exactly what customers need and want, developing new models of customer behaviour.
- Continually update fraud models to evolve with customer behaviour, ensuring they stay ahead of fraudsters.
Data science in action
Though far less than one percent of all transactions are fraudulent and most banks already stop a large majority of that fraud, the losses for the banking industry overall still total many millions of dollars annually, so there’s more work to do.
Let’s take a look at a case of a bank that, using their current technology, stops about 80 percent of fraudulent transactions (which is around the industry standard), meaning they are still losing millions of dollars on the fraudulent transactions they don’t catch. They want to do better and raise that number to at least 90 percent, but without increasing staffing or spending significantly more time - a perfect opportunity for big data.
The bank’s data science team looks to the huge amount of data at their disposal, but each dataset is siloed and thus difficult to analyse individually, much less analyse as a whole. So their first step is to combine data from all possible sources to get a full picture from which they can identify risky transactions. With a complete dataset, they are then able to start developing and testing different models to determine which combination or set of signals is best able to automatically predict fraudulent transactions.
Once the model is perfected, the bank’s data science team is able to deploy it to production. Working with real-time data, the model is able to not only effectively identify more fraud faster, but continually learn to detect new and emerging patterns. By embracing data science and machine learning, the team is able to reduce losses and keep up with new risks that would have been impossible by simply statically looking at historical data.
While we can’t say that fraudulent activity could have been completely stopped in the Tesco Bank case, with a similar model, it’s possible that it could have been detected and stopped sooner before affecting the business significantly and thousands of customers.
The bottom line
Whether in marketing, risk management, product design, finance, actuarial science, underwriting, or claim management, staff should be empowered to use data to drive the business forward. With everyone on board, your business can start to seize the analytics opportunity and join the big data revolution and the age of algorithms.
It’s not a good idea to wait until something detrimental happens to put the right systems in place to more accurately and quickly detect fraud - the time is now to ensure data science systems and teams are in place that can access and easily analyse data in real-time to detect the latest trends and anomalies.
Florian Douetteau, data scientist and the co-founder and CEO, Dataiku