United StatesFinance and insurance2017 – ongoingRemitly
Machine learning fraud models for money transfers
Remitly said its fraud models cut the intervention rate by over 25% and transaction losses by over 30% in 2023
Problem
Fraud checks on money transfers can delay legitimate senders. Remitly wanted to stop identity theft and card fraud while intervening on fewer genuine transactions.
Approach
Remitly has used machine learning against fraud since 2017. Its engineering teams run models that review tens of thousands of transactions per hour and look for patterns tied to identity theft and credit card fraud.
Results
Takeaway
Remitly tracked fraud losses and the rate of interventions on customers together, since a tighter check can raise one while cutting the other.
Categories
- Industry
- Finance and insurance
- Business function
- Finance, risk and legal
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Two Remitly executives wrote that its machine learning models review tens of thousands of transactions per hour. They gave year-over-year falls in intervention and loss rates for 2023 against 2022. No absolute numbers were given.
Full details are in the original sources.