Bust-out fraud detection: why trajectory scoring beats snapshot metrics

Short summary

Bust-out fraud is first-party fraud where a borrower builds a clean credit history to get limit increases, then draws the entire line and disappears. The challenge is distinguishing this from genuine financial distress, which produces similar utilisation and delinquency readings. The author demonstrates that trajectory-based features (utilisation slope, payment streak patterns, limit growth, cash-advance share) separate the two far better than snapshot metrics like current utilisation.

• •Bust-out fraud hides by looking like a model borrower before drawing the full line and defaulting
• •Snapshot-based scoring (utilisation, DPD) misclassifies ~70% of flagged accounts as genuine distress, not fraud
• •Trajectory features like utilisation slope, recent full-payment streaks, and limit growth correctly separate bust-out from distress

Generated with AI, which can make mistakes.