At Ravelin’s core is an machine learning system that evaluates each customer’s risk level in real-time. It uses hundreds of signals about each of your customers and taps into data across the entire Ravelin network to predict whether a customer is likely to be fraudulent. It outputs a fraud score to represent this prediction. Machine learning is simply a form of artificial intelligence that enables computers to learn without being explicitly programmed. It’s especially good at recognising patterns in data and therefore equally good at spotting anomalies in those patterns. This makes it a great approach for preventing fraud. Benefits of machine learning include:
- highly adaptive and self-learning, meaning minimal maintenance is required compared to other approaches.
- work extremely well in high volume and peak-scale environments, indeed they improve with more data.
- provide instant and constantly evolving scores based on events in a customer journey so score is always current.
- predictive capabilities mean frauds can be declined pre-checkout to avoid chargebacks.
Ravelin creates networks by connecting customers that share any of the following:
- False positive: When an a legitimate order is prevented by Ravelin. Usually this is spotted when a customer contacts your support team.
- False negative: When a fraudulent order should have been prevented by Ravelin, but is not. Usually this is spotted when a chargeback is reported.
- Fraud score: The probability that a customer is fraudulent.
- Score distribution: The number of customers getting each possible fraud score.
Improving performance with manual review
Human feedback is critical to helping machine learning models learn and improve. Your manual review teaches your machine learning models how to better identify fraudsters and helps improve Ravelin's accuracy.