This guide explains Omniscore—a comprehensive safety rating designed to help you distinguish between legitimate customers and fraudulent activity.
Omniscore is a transaction reliability rating. It uses advanced Artificial Intelligence (AI) to analyze billions of historical data points to predict the safety of a new order in real-time.
Instead of evaluating data points in isolation, Omniscore aggregates historical patterns, current behavior, and identity clusters into a single numerical value. This allows your team to make rapid, data-driven decisions—Approve, Decline, or Review—with minimal manual intervention.
Omniscore is powered by two distinct types of Machine Learning (ML) that work together to provide a holistic view of risk.
-
Supervised Machine Learning (The "Veteran Analyst")
-
Function: This engine is trained on years of historical outcomes, including confirmed fraud, chargebacks, and legitimate transactions.
-
Purpose: It recognizes "known" fraud patterns and established behaviors that have historically led to loss.
-
-
Unsupervised Machine Learning (The "Instinct")
-
Function: This engine analyzes live data to find hidden relationships and anomalies without needing prior labels.
-
Purpose: It identifies "unknown" threats—newly emerging fraud tactics or sophisticated bot attacks—by spotting suspicious deviations from normal behavior.
-
Omniscore ranges from 0.1 (Highest Risk) to 99.9 (Lowest Risk). To simplify decision-making, these scores can be categorized into risk levels.
Omniscore is a predictive tool, not a static rule. You can integrate it into your workflow to automate responses based on your business’s specific risk appetite.
By using Omniscore as a primary filter, you can simplify your rule logic. Choose a threshold that aligns with your tolerance for risk:
-
Aggressive Protection: Decline orders with a score < 61 (targets ~5% of total volume).
-
Balanced Approach: Decline orders with a score < 25 (targets ~2% of total volume).
-
Conservative Filter: Decline orders with a score < 13 (targets ~1% of total volume).
For teams new to AI-driven scoring, we recommend a "No Change" rule. This allows Omniscore to run in the background without affecting live orders. After a 60–90 day period, you can audit the results against actual performance data to see exactly how many chargebacks the score would have prevented.
For Analysts: Prioritize high-value manual reviews by focusing on "Mixed" scores, while automating the clear "Safe" and "Risky" orders.
For Business Leaders: Improve the bottom line by reducing "False Positives" (good customers being blocked) and lowering operational costs associated with manual review and chargeback management.
Comments
0 comments
Article is closed for comments.