- Global financial service provider offering secure methods for transferring money
- Instant access to money for online payments
- Real-time processing of payments and transactions
- Fraud had to be prevented before it occurred, not just detected
- Sent historical data to accurately train Sift's model from the onset
- Used Sift to automate fraud decision-making
- Reduced full-time fraud team to 0
- Increased sign-up to usage conversion rate of good users by 15%
Easier, safer online payments
Entropay offers users a quick and easy solution to online payments. The company allows its users to instantly access funds via a prepaid virtual VISA card that can be used to make purchases anywhere online. While Entropay’s innovative solution greatly simplifies the process of online payments for its users, the ease and immediacy of its prepaid cards also make it incredibly attractive targets for fraudsters. From the very beginning, Entropay was prepared for the possibility of fraud. With a payment product that was simple, anonymous, and instantaneous, fraud was all but inevitable. Entropay’s team initially built and maintained a proprietary fraud rules engine, which would flag users and transactions for manual review. Due to the scale and immense amount of time needed for the manual review process, Entropay had a dedicated team of 8 fraud analysts reviewing transactions around the clock and a data analyst managing its rules system. Because its rules weren’t updated in real-time and had to be tweaked manually, Entropay’s system was always playing catch-up and couldn’t stop fraud as it was happening.
Stopping fraud before it happens
Due to the immediacy of Entropay’s service, every lost dollar was unrecoverable. Over 90% of the fraud that Entropay faced came from stolen credit cards. Fraudsters would use stolen credit cards to fund Entropay’s virtual Visa cards, which the fraudster could then spend immediately. Not only did Entropay have to worry about the financial loss and the chargeback fees, Entropay’s team also had to keep their chargeback rate within their bank’s threshold to avoid risking additional fees or the termination of their account.
Because Entropay’s product was expected to process immediately in real time, its fraud had to be prevented before it occurred, not just detected.
When Mark Anthony Spiteri, the Head of Product and Business Operations at Entropay, came across Sift, he was impressed by its innovative approach to fighting fraud using powerful large-scale machine learning. He was excited that Sift looked at thousands of different fraud signals, as he knew that it would be very difficult to capture and analyze all of that information with Entropay’s own internal systems.
“Fraudsters aren’t going to be using the same patterns, they’re going to be moving on - we were always playing catch-up.”
Automating fraud prevention and improving customer conversion
The Entropay team was able to quickly implement Sift over a single weekend. By uploading a large amount of historical data with marked good and bad transactions, they efficiently trained their customized machine-learning model, tailoring it to Entropay’s business and needs from the very beginning. Initially, Entropay trialed Sift for just two of its markets to test its effectiveness. For both markets, Sift not only helped Entropay block more fraudulent orders but also improved its conversion rates for good customers. After just two months, Sift was so effective and accurate that Entropay was able to switch to using Sift for fighting fraud instead of relying on their internal rules engine.
Today, Entropay’s fraud management system is fully integrated with Sift, and it uses Sift to automate order approval and cancellation decisions. Their team trusts the predictive accuracy of the Sift Score, using that number to automatically determine whether to accept, block, or further challenge orders. Additionally, the Entropay team will often use the Signals and Score History features to understand why Sift predicted a particular user to be fraudulent, informing their next actions.
“Sift analyzes thousands of signals to detect fraudulent behavior which would’ve been impossible for us to do ourselves.”
Incredible accuracy with exceptional results
Sift’s extremely accurate fraud detection allowed Entropay to automate its fraud process and reduce the hours spent on fraud management by 90%. Entropay transitioned its team of 8 full-time fraud analysts from focusing on stopping fraud to concentrating on customer success and support, effectively reducing its fraud team to zero. By spending more time on growing its business and leaving fraud to Sift, Entropay better supports its users and has increased its conversion rate of good users by 15%.
The incredible customization and immediacy of Sift’s machine learning results make it a possibility for Entropay to not only stop fraud but also proactively grow its business.