Data science in finance: key applications
Data science has seen enormous growth over the past few years in numerous industries. It regularly ranks as one of the top career paths in the US and is only expected to grow considerably in the years to come, with the Bureau of Labor Statistics predicting more than 30% growth in the field between 2019 and 2029.
The finance industry was an early adopter when it came to using data science to improve profitability and reduce costs and risks – one significant example being the initial work in 1956 by engineer Bill Fair and mathematician Earl Isaac to create what would be refined to become the FICO credit score.
So what are some of the most significant ways that data science is used in finance today?
Risk Analytics & Financial Scoring
Even now, more than 65 years after work began on creating a system for scoring the creditworthiness of consumers, the credit score is still relevant and widely used. Consumers are familiar with the system and tend to have at least a basic understanding of what factors influence their score and how it can be improved.
With traditional scoring models, there are simple flaws that can cost banks. In some cases, high net worth individuals may have lower than expected credit scores because net worth and income are not factored highly into the calculation. On the other hand, credit card churners may be able to maintain high scores, while being low-value customers to the banks.
By bringing AI into the process, financial institutions can utilize not only historical data, but make predictions based on income and earning potential. This allows institutions to tap into a potential pool of customers that is typically overlooked by traditional scoring models. Not only that, but AI offers the ability to make these decisions faster than ever.
Until recently, most data used in the finance industry had been collected and then processed in batches. This means that while data can still be recent, the delay in processing means there is still room for costly issues to arise – particularly when it comes to fraud.
Thanks to the broader availability of the required computing power, and the accompanying lower cost, real-time analytics is now available to financial institutions, and many are using it in a number of ways such as:
- Fraud Prevention.
Fraud costs the finance industry billions of dollars each year, and the majority is linked to credit card transactions. In 2018 alone, the cost of this fraud was over $9B dollars, so improving fraud detection is a huge priority. New real-time analytics technology allows for risk analysis to happen at the time a transaction is processed allowing for near zero-latency fraud prevention.
- Critical Application and Service Monitoring.
Network uptime and platform security are critical for the financial industry, and using real-time analytics for Network Behavior Analysis (NBA) is a useful application of the technology. At the basic level, unfiltered network data is converted into metadata which can then be combined with machine learning models to identify network anomalies in real-time. These anomalies, whether security- or performance-related, can then be reported immediately to the relevant teams for review, or even dealt with autonomously.
- Improving Customer Experience.
Real-time analytics paired with customer relationship management software can significantly improve the customer experience. In an age when customers can interact with their financial institutions in a number of ways, including apps, websites, call centers and in-person, there is an expectation of a fast, seamless and personalized experience. In a 2019 survey, primarily focused on banking and financial services organizations, 60% of respondents replied that real-time analytics was extremely important for the performance of their organizations.
The value of customer satisfaction should not be underestimated. A McKinsey retail bank customer survey from 2018 revealed that deposits at banks with high levels of customer satisfaction grew 84% faster than those with the lowest scores.
In addition to improving the customer experience, identifying and retaining good customers is a top priority for financial institutions. By combining customer data with AI, financial institutions can better predict how a customer will act and put together models to identify who are likely to be the highest value customers, and which customers should be dropped.
Algorithmic trading, the practice of using software automation to make trades on stock markets has seen huge growth over the past decade. As much as 70-80% of all trades in the United States are believed to be completed using this method. There are good reasons for this too; it is faster, has a history of success, and computers are available 24/7 whenever the markets are open around the world.
This method of trading is not without risk, as evidenced by the Knight Capital incident that occurred on August 1, 2012. On that day, software run by the firm, which was believed to be inactive, went rogue for around 45 minutes at a cost of approximately $10M per minute.
Even with the risks involved, algorithmic trading is growing steadily, with a predicted compound annual growth rate of over 11% for the next five years. Institutions are now focusing on bringing machine learning into the process, one of the most prominent examples being JP Morgan using reinforcement learning algorithms in their futures market trades, with algorithmic trading now making up to 20% of their total trade volume in this area.
The growth of data science in finance offers a win-win for both financial institutions and their customers.
For institutions, it offers the ability to reduce risk by identifying undesirable customers, reduce the risk of fraud, and increase success with stock market trading through automated high-speed trading.
On the other hand, customers can benefit from improved and highly personalized customer service, and in some cases, the ability to access credit or investment opportunities for which they may have been previously overlooked.