Discovering the fraud bottleneck in intelligent document processing (IDP)
Key to Payoneer’s success has been an intense focus on operational efficiency and the constant prioritization of their customer experience. To meet the challenges of staggering growth across a wide diversity of compliance regimes, Payoneer sought the help of various technologies specializing in document processing.
“We were looking for a way to introduce technology that would simplify our document intake process... In that journey, we quickly realized that reviewing the documents for fraud was a prerequisite, because document processing technology might work to lift the content from the document, but it has no idea if a document has been modified.”
They were looking at proof of address, income, and certificates of incorporation documents coming from a huge variety of markets and submitted in a host of different languages and formats: PDFs, images, scans, screenshots, etc.
Still, they have high standards and performance that must be maintained. Customer approval representatives work with regional regulatory experts as their back up and adhere to strict service-level agreements to deliver fast turnaround times.
Trying to detect digital forgeries which cannot be seen by the naked eye under such conditions was creating unacceptable levels of doubt and risk in the onboarding process.
Worse yet, like many other digital-first financial services Payoneer encountered a new type of fraud in their submissions: mass, serialized fraud, where single documents were being used as templates to spawn dozens or even hundreds of forgeries.
“With the volumes we process, it is critical for us to introduce robust technology that supports our document fraud checks.”
Uncovering serialized fraud
Resistant AI embarked on a true partnership with Payoneer, developing multiple cutting-edge detection capabilities to help overcome fraud challenges.
First, forensic detections were applied and tested against large batches of documents submitted by customers. Because Resistant AI’s Document Forensics capabilities don’t require reading the content of the documents to detect forgery, it was particularly well suited to work across the large variety of languages Payoneer deals with. It also provided customer approval reps with detailed findings.
Then, serial fraud detection capabilities were deployed to compare all documents across completely different submissions, stopping hundreds of serial fraud attempts per day.
Confident and accurate decisions
Hyper-granular verdicts were customized to meet Payoneer’s specific risk appetites with 99.2% accuracy, triggering various workflows for document acceptance and declines to speed up the work of customer approval reps.
When documents were escalated for investigation, Resistant AI learned and adapted from them, continuously reducing the amount of documents going through manual fraud review.
Now, with more than 82% of their document fraud decision-making supported by Resistant AI and only edge cases escalated for review, Payoneer can confidently continue its journey to accelerate its entire customer onboarding process.
“What impresses me most is the fast reaction. After we raised issues, Resistant AI took only 24 hours to develop new logic which gave us a precision rate we could only dream of... Probably the best tool we have in our review flow.”
Verify any document for fraud
