Risk score reasons are a set of data that provide you with specific and understandable reasons for why a risk score is high or low. This data is exclusive to the minFraud Factors service. Learn more about risk scores.
The risk reasons output codes and reasons are currently in beta and are subject to change. These outputs are being actively developed and tested, and may undergo modifications that could impact their structure, format, and content. While we strive to maintain stability, we recommend that you use these beta outputs with caution and avoid relying on them for critical applications. Your feedback is valuable and will help us improve the final release.
Accessing minFraud Risk Score Reasons
Risk score reasons are present in the API response for queries sent to the minFraud Factors service.
You can also view the risk score reasons for any transaction on the Transaction Details page in the Admin portal for transactions scored with the minFraud Factors service.
- Sign in to your MaxMind account and navigate to the minFraud Transactions page (direct link, login required).
- Click the transaction ID to open the Transaction Details page.
We return a risk multiplier value in addition to a reason code and a human-readable string. Not all risk score reasons are returned—only risk score reasons that change the risk score significantly for a given transaction are returned. Multipliers greater than 1.5 and less than 0.66 are considered significant and lead to risk reason(s) being present.
Recommended use cases
Risk score reasons provide fraud teams with the data needed to deeply understand risk patterns and make informed decisions.
- Use the data for forensic investigation. If you question the accuracy of a score for any reason, the risk score reasons can give you an understanding of what our model is doing. The multipliers give you a sense of the relative magnitude of each reason, and how they impact the overall score.
- Use the data for post-incident analysis. Take the mean multiplier of risk score reasons to get insight into what is driving the score for a given snapshot of time. You can analyze a specific target class, such as false negatives or false positives.
- Use the data for pattern analysis over time. Fraud patterns change constantly, so you can identify which patterns are driving the scores and fine-tune specific aspects of your fraud strategy accordingly.
- Use the data to feed your own machine learning models. Surface new features that add lift to your ML models with over 30 new data points that we consider high signal for fraud and risk applications.
Feedback
We would love your feedback on how we can improve this feature to help you with investigations, post-incident and pattern analysis, and ML optimization use cases. Reach out to us at customersuccess@maxmind.com to schedule a feedback session.
This page was last updated on .