Fraud prevention is a balancing act. Although merchants should keep fraud rates low, it is imperative that good customers are not blocked-in error – An issue known as False Positives.
If merchants limit themselves to only certain metrics, it is not possible to see the full picture. And this could be costing businesses a significant amount of revenue in the long run. Financial advisory firm, Aite Group, estimates that false positives could lead to e-commerce losses of $443 billion by 2022.
However, by tracking too many indicators, more time is spent on time and resources to measuring fraud rather than acting on it. Fraugster’s team of analysts and experts recommend the following seven key performance indicators for fraud prevention success:
Fraud rate
The Fraud rate is a calculation based on known fraud cases only.
Only approved transactions that turned out to be fraudulent are counted. Whereas, declined transactions are not defined as fraudulent transactions and therefore not included in the fraud rate calculation.
Incoming Pressure
Incoming Pressure is the percentage of attempted transactions that turn out to be fraudulent.
This KPI is closely connected to the Fraud Rate and calculated retrospectively, by using chargebacks or other ways to tag transactions, such as linking or manual review. A best practice tip is to make sure that a fraud prevention solution measures incoming pressure to include fraud out of the total population as well as all declined transactions.
Final Approval Rate
The Final Approval Rate is calculated based on the status of transactions regardless of a fraud prevention solution’s decision.
All transactions that received the Approved status are considered when calculating your final approval rate.
This fraud KPI is the simplest to measure. While it does provide a good macro indication of how a fraud prevention solution is performing, without the other metrics, there is a limit to the decisions that can be made.
For instance, if only focused on the Final Approval Rate in isolation, it is easy to miss the relationship that it has with the Incoming Pressure. Is an 80% Final Approval Rate good enough? If there is no incoming pressure, then the answer is no. If there is a spike in incoming pressure, then this could be an acceptable rate.
It is important to note that the fraud prevention solution will not be the sole gatekeeper when it comes to declined transactions. Third parties provide additional layers of security as well. For example, 3DS2 could request further verification from customers. And the card issuer will also be responsible for declining a certain number of transactions. A good fraud prevention solution should align strategies with the Issuer to ensure optimal performance.
Precision
Precision is the percentage of fraudulent transactions from the total number of declined transactions.
Considered as a derivative of the false-positive ratio, Precision measures the accuracy of a decision. To measure, the population that has been declined should be observed. For instance, if you were to decline 100 transactions, how many of them were bad?
It is important to note that the number of fraud cases that have slipped through undetected (or the number of legitimate transactions that were blocked by mistake) will adversely affect this KPI.
Recall
Also known as the Catch-Rate, Recall is the percentage of fraudulent transactions declined from the total number of transactions tagged as fraudulent.
The more fraudulent transactions that are identified and blocked, the higher the Precision will be. For the best performance, Recall should be as high and precise as possible.
An AI-enabled process, Data enrichment is one of the most accurate ways to improve recall. Utilising the most basic purchase information, such as email address and first name, and expanding them into thousands of additional data points, an AI, such as Fraugster, can use this richer information to accurately score transactional risk, and make an accurate approval or decline decision.
Decline rate
The Decline Rate is the percentage of transactions declined from the total number of transactions.
Whilst this metric provides the number of transactions blocked, it can be misleading without context. For this fraud detection indicator to truly have merit, an accurate reading of the Recall and Precision metrics must be considered. Once there is an accurate record of both of these indicators, it is possible to understand how many transactions are bad and how many were blocked.
Understanding why a transaction has been declined is vital to weeding out false positives, boosting approvals, and catching more fraud. Therefore explainable AI is an extremely important aspect of fraud prevention KPI success.
Fraugster provides an AI Score that not only provides a highly accurate determination of the likelihood of fraud, but each decision comes with a full explanation as well:
Good User Approval Rate
The Good User Approval Rate is the percentage of good transactions that were legitimate from the total number of approved transactions.
In a nutshell, merchants want to maximize legitimate transactions. Therefore it is critical to track this specific metric independently.
A merchant tracking the Good User Approval Rate and comparing it with Incoming Fraud Pressure, as well as the overall Approval Rate, will provide a good indicator of the performance of their fraud prevention tools and strategy.
Make better business decisions by measuring what matters. With the right tools, the most relevant information can be aggregated to make the best decisions possible.
Editors note: This post was contributed by Fraugster