Real-Time Card Payment Fraud Detection
The Security and Fraud Director needs to block fraudulent card transactions before authorising the payment, analysing hidden patterns among millions of legitimate operations.
Where the money comes from
What to do now
- 1Quantify the cost of a fraud and that of a wrongful block.
- 2Train the model with the labelled transactions and measure with recent data.
- 3Activate it in silent mode (scores only) for 4 weeks.
For your committee
One page with the decision, the impact, the risks, the steps and the KPI to track.
The print dialog opens: choose "Save as PDF".
Watch out: more volume is not always more profit
Blocking aggressively reduces fraud but also the revenue and loyalty of good customers. The threshold is set by comparing the fraud avoided with the sales and customers lost.
KPI for the committee
Fraud loss (basis points on volume) and false block rate.
Detail for your data team
Proposed technique: XGBoost
Why: Fraud presents highly non-linear and imbalanced relationships (less than 0.5% of operations are fraud). XGBoost stands out for its ability to detect these subtle combinations with very high accuracy.
Cost of getting it wrong: Letting fraud through costs money directly; blocking legitimate customers destroys the user experience and repeat usage.
Data needed: Payment Switch: Amount, merchant, country, currency, exact time and authentication method.; Cardholder History: Usual average spend, frequent locations and speed between consecutive transactions.