In our last blog, we talked about the informal poll we conducted. We asked a LinkedIn group of credit professionals what their biggest fears as a risk manager were. We also took a deeper dive into fear #1: Missing default-indicating red flags (with 43% of the votes). If you haven’t already read that blog, you can do so here. In part II, we will take a closer look at fears number 2 and 3. In our poll, 34% of our voters said that being misled by false positives and negatives was a cause for concern. And 23% chose an inability to predict future risk as their third-biggest fear. Here are the results once again:
Without further ado, let’s get down to the nitty-gritty of it all!
Fear #2: False Positives and False Negatives – A necessary evil or avoidable?
With default loan rates projected to increase through 2022, financiers are, understandably, trying to tighten their risk mitigation methods to protect their profits and their investors. But as with most things in life, balance is key here. If you tighten the default-indicating screws too much, you run the risk of increasing false positives (when your risk screening wrongly flags a borrower as risky). The opposite is even worse. Loosening the screws too much can increase false negatives (when your risk screening wrongly marks a risky entity as safe). Both are scenarios that risk managers could do without.
The problem with false positives and negatives
False positives are probably the bane of a credit risk manager’s existence. Checking out an alert requires extra time and, sometimes, an extra set of eyes to conduct a deeper analysis of the flagged borrower’s financials. It is a time-consuming process, one that is ultimately inefficient if the additional investigation proves nothing. And it is a costly process too. To deal with their caseload of alerts, financiers need to add more trained analysts to their team, resulting in a bloated risk-mitigation department and a hefty bill. According to Global Trade Review, banks waste over $3 billion each year chasing down false positives with AML (anti-money laundering) screening alone. What’s worse is that dealing with false alarms day in, and day out can result in alarm fatigue. This can cause managers to falsely assume that the next alert is a red herring too, making them throw the baby out with the bathwater.
While a false positive is an irritating waste of time and money, false negatives are inherently more treacherous. For example, a false positive Covid test result will warrant that you undergo quarantine unnecessarily, which is an inconvenience for sure. On the other hand, receiving a false negative Covid test result is far more dangerous. A negative test result gives you the freedom to be out and about, potentially allowing the virus to spread to others. Similarly, false negatives leave your organization vulnerable to serious financial and reputational hits, some of which are difficult to recover from.
In addition, both false positives and negatives make it tougher to set accurate parameters or benchmarks of what constitutes true default-indicating behavior. So, you are stuck in an endless loop of setting and reconfiguring your baselines over and over again.
3 ways to reduce false positives and negatives
Financiers accept false positives and negatives as an inevitable part of the risk mitigation process. While that is true to a certain extent, you should, at the very least, be able to completely eliminate false negatives. Thankfully, there are ways to get you closer to that elusive 100% accuracy rate and downgrade false positives and negatives from being a necessary evil to an avoidable one.
1. Improve the quality and quantity of your data
To reduce false events, you must first improve the quality and quantity of your data points. The efficiency of your risk mitigation system is irrevocably tied to your ability to see everything that is going on with your borrowers. For this holistic view, you need lots and lots of data. Simply put, the more comprehensive and exhaustive your data is, the more precise your risk monitoring can be. In other words, your risk monitoring is only as good as your data! In addition, the data you use should be current. Using outdated data can, obviously, lead to inaccurate results. Keeping the information fresh, on the other hand, ensures a clearer picture and real-time risk monitoring. More importantly, it allows you to catch newer patterns of risky behavior.
2. Streamline your workflow with automation
A reliable way to reduce the white noise, especially when you are dealing with a large portfolio, is to streamline your risk-mitigating process with automation. Collecting and analyzing all the information needed for accurate results can seem like an insurmountable task. Automating time-consuming processes such as data collection and organization can, therefore, streamline your credit monitoring workflow significantly. As an added benefit, automation gets rid of human errors that can taint the process and frees you up to do other important things.
3. Switch from legacy systems to AI-augmented systems
Traditional legacy systems that rely mainly on historical data and manual analysis to flag default-indicating red flags take a linear approach to credit risk monitoring that is incomplete. The resulting gaps in monitoring are what lead to the dreaded false results. Conversely, AI-augmented systems such as TRaiCE use advanced analytics and have a non-linear setup that makes for a more complete process.
One of the biggest benefits of an augmented system is its ability to process and investigate significantly more information than any portfolio manager or analyst ever could. These powerful systems can also discover the nexus or connection between seemingly disparate events, spotting patterns that are invisible even to the great Sherlock Holmes!
Subjective, manual risk controls are unreliable and can create a false sense of security or of panic (depending on which side of the risk/safety equation the controls are skewed towards). To reduce false results, traditional methodologies should be reinforced with AI-augmented systems. Only then can financial institutions cover all the gaps in monitoring that currently exist. Without this, they will be forced to keep adding costly, yet ineffective, layers of protection.
Now onto fear #3:
Fear #3: Predicting future risks
Predicting future risks – A hard but necessary exercise
Hindsight, as they say, is 20/20. Once the dust has settled, it is easy to look back and realize what the right course of action was. Foresight, on the other hand, is a different kettle of fish altogether. Because no one can really see into the future, foreshadowing cataclysmic events and creating contingency plans for them is an infinitely harder task. As US Admiral Charles Nimitz put it, hindsight is notably cleverer than foresight.
But predicting future risk, hard as it might be, is a necessary exercise for financial institutions everywhere. Without it, financiers are reduced to playing a reactive game of whack-a-mole, instead of being proactive with their risk mitigation. Often, this can prove to be costly. According to McKinsey, using improved default prediction systems can help save banks $10 million annually.
Of course, the process of expecting the unexpected is not an exact science. You must use whatever information you have at present to forecast the likelihood of credit risk in the future. So, there are limitations, and not every prediction will be on point. That said, the advantages far outweigh the limitations. Proactively foreshadowing loan defaults allows you to be prepared for whatever may come your way. It allows you to always be in the driver’s seat. Conversely, reactively scrambling to limit the damage leaves you at the mercy of others. In addition, forward-focused risk monitoring can give you a competitive edge over others and can act as a springboard for faster, more timely remedial actions.
How predictive analytics is changing the game
While 100% accuracy in predicting future risk is likely to remain a pipedream, recent advances in technology have greatly improved the precision with which it can be done. Traditionally, financiers predicted future trends using historical data and other standard financial metrics such as credit scores. Relying only on this small dataset to inform the future provides a hazy and incomplete picture. To paint a clearer picture, a whole lot more data and a more dynamic approach are needed.
Today, there’s a huge amount of data out there waiting to be analyzed. The problem is that a large chunk of this information is unorganized and, therefore, nonsensical in nature. Predictive analytics solves this problem by using sophisticated machine learning techniques that can leverage this huge dataset by organizing and analyzing it. What’s also great is that these powerful algorithms can learn and re-configure themselves as parameters change, making them adept at identifying newer patterns of risk.
How TRaiCE eases a portfolio manager’s fears
TRaiCE’s algorithms leverage big data. The system combines traditional data points with data from external sources such as newsfeeds, social media, and stock markets. Including these richer datasets improves the accuracy of its predictions and gives you a customized early-warning system that can predict individual client risk for the next 3 to 6 months. What this means is that default-prone customers are flagged sooner and with more accuracy. You can then take the necessary remedial steps needed to protect your profits, with time to spare. In other words, TRaiCE allows you to be proactive with your risk monitoring instead of 10 steps behind. This will surely also go a long way in easing the fears that a lot of portfolio managers have to deal with daily.