Now that we’re into International Fraud Awareness Week, I find the amount of discussion around the role of data in fraud defense capabilities striking. It’s a great focus area, especially as machine learning and advanced data analytics become more accurate in predicting fraudulent patterns for banks as well as corporates. But it’s a week about awareness, right? As such I think we need to be more aware of the data availability and data quality that will determine the success of an anti-fraud initiative.
The amount of data available to banks of any size, and to most larger corporates is simply staggering. I’ll wager that most senior executives haven’t heard much about “zettabytes.” It’s a metric that is calculated by using the formula 10 to the 21st power. That’s one followed by 22 zeroes. At the beginning of 2020, according to one study, there were 44 zettabytes of data generated from digital sources in the world. That number will triple by 2025. Managing this is a challenge, to say the least, and insuring its quality at your bank or corporate is a competitive point of the first order.
Data is not a monster looking to overwhelm your anti-fraud efforts. It is, with the right partner, people, and analytics, the energy that determines the power of your anti-fraud efforts. You can create the most advanced predictive analytics and algorithms, but if the quality of the data is lacking, predictive analytics won't help. Updating and constantly optimizing that data is essential, especially for banks, and in my experience, it is not always at the level it needs to be. You don’t want an unreliable fraud alert. You don’t want to misinterpret a spending pattern that could indicate fraud. Quite simply, bad data can be more expensive than a solid data analytics solution when it comes to fraud.
I’m encouraged to see financial institutions looking at the quality of data, as well as what happens during its journey from customer to database. Banks are asking the right questions: Where did this data originate from? Are the fraud analysts looking at the same datasets as the rest of the company? Has the data been transformed through an earlier process? Do I need to interpret the data differently? That’s how data squares with anti-fraud efforts. We need to understand what is happening with the data before it reaches the fraud monitoring tool. This issue goes into a very broad area of data quality and analytics as it usually involves data transformations (i.e. making the data ready to be handled by systems), but that same data is also used to develop predictive models for fraud detection. So the concern is having reliable data that we can be confident in for an anti-fraud solution.
If you do have reliable data it can actually enhance the customer experience.
Your first reaction may be “how can an enforcement tool meant to catch criminals be a positive customer experience?” Good question. It comes back to the predictive analytics mentioned earlier. If you're using predictive analytics, it calculates risk scores for every transaction. This approach enables FIs to focus on the highest risk scores and their associated transactions. Transactions with the highest risk scores are, for example, automatically assigned to the most senior fraud investigators. But that same risk score can also be used to improve the customer experience. It enables banks to apply less restrictive fraud controls on less-risky customer behavior, reducing the volume of investigations and payment holds. Data analytics and AI have significantly improved the ability to catch fraud patterns, and even head them off before they happen. Data, done right, is a reliable fraud defense.
A few words about predictive analytics as it relates to this mix. It's one of the detection capabilities out there, but it's not the only one. It’s seen as a silver bullet by many organizations, but based on my experience and interactions with FIs, AI still needs to be combined with rule-based analytics and behavioral profiling. It's the combination of detection capabilities that complement each other. In this case, the whole is greater than the sum of its parts. So, detection capabilities should overlap each other a little bit in order to close the gaps in detection technology. That's why we promote a hybrid approach or layered protection strategy. In my experience, a hybrid approach has better detection rates and fewer false alarms, which cuts costs on investigations. More importantly, with fewer false-positive alarms you will not disrupt the customer experience.
The quick wins in this space right now revolve around the customer experience. The objective of data usage is to catch the fraudsters. But if you frustrate customers in that payment process, they go to the competitor on the other side of the street, because that's easier. Catch the bad guys, reward the good guys. Both can be done.
For further insight into the payments and banking industries, subscribe now and stay up-to-date on the latest tips, trends, and topics. You can also check out The Payments Podcast, where experts engage each other on the real-world factors impacting your industry.