Fraud Detection
Stop fraud before clearing
AI fraud scoring runs in the clearing pipeline. No separate API. No added latency for the end customer.
Fraud model
How fraud scoring works
Before the clearing instruction is sent, each transaction receives a score from enza's fraud engine. The model assesses velocity patterns, geographic signals, device fingerprint proxies, and amount anomalies. Transactions above a configurable risk threshold are flagged, declined, or queued for manual review.
- Velocity analysis across card and account signals
- Geographic and corridor anomaly checks
- Thresholds configurable by product and corridor
Clearing window
The clearing window is the critical moment
If fraud is caught after clearing, you own the chargeback. enza's model runs inside the clearing window, typically 40-80ms, so every intervention occurs before money moves.
Accuracy metrics
Precision from African transaction data
Our model is trained and benchmarked on African card transaction patterns, not retrofitted from a global model. Chargeback rates in early-access pilots stayed below 0.05%.
From our early-access program with fintech partners in Nigeria and Ghana.
98%
Fraud precision, internal benchmarks
<0.05%
Chargeback rate, early-access portfolios
No extra integration
Fraud detection comes with the API
Connect to enza for card acceptance and fraud detection is already in the pipeline. There is no separate fraud API or additional webhook setup. Configure thresholds in the dashboard.
Request API accessStop fraud before chargeback
Connect your card product to enza for real-time fraud scoring from day one.