NovaPay is testing a new fraud rule that blocks card transactions when a payment comes from a new device and the transaction amount is above $180. The risk team wants to know whether the rule reduces fraud losses without creating too much friction for legitimate customers.
The team ran the rule for 14 days on 1,200,000 transactions and compared outcomes to the previous policy.
| Metric | Baseline Policy | New Rule | Change |
|---|---|---|---|
| Fraud rate | 0.80% | 0.80% | 0.00 pp |
| Precision of blocked transactions | 0.41 | 0.36 | -0.05 |
| Recall on fraud | 0.52 | 0.68 | +0.16 |
| F1 score | 0.46 | 0.47 | +0.01 |
| False positive rate | 0.93% | 1.74% | +0.81 pp |
| Fraud dollars prevented | $468,000 | $612,000 | +$144,000 |
| Legitimate customers blocked | 11,070 | 20,736 | +9,666 |
| Manual review volume/day | 2,900 | 4,250 | +1,350 |
| Customer support contacts/day | 410 | 690 | +280 |
The new rule catches more fraud, but it also blocks many more good customers. You need to determine whether the tradeoff is acceptable and how to measure the rule's net business value.