Banking fraud prevention: stopping fraud before settlement

Fraud costs a bank twice: once in stolen funds, and again in genuine customers who are wrongly blocked and take their business elsewhere. This digital bank faced rising account takeover and payment fraud, while its existing controls were also stopping too many legitimate customers. It needed to catch more fraud without making life harder for everyone else.

What was going wrong

Account takeover (ATO) is when a criminal gains control of a real customer’s account, usually with stolen login details, and moves money out. The bank was seeing spikes in ATO and payment fraud. Its rules engine, a fixed set of if-then checks, could not keep up as criminals changed their patterns. At the same time, it was flagging too many normal transactions as suspicious.

How the bank used Snipeyes FD

The bank deployed Snipeyes FD, our real-time fraud detection product. FD scores every login and payment as it happens. It looks at how the customer behaves, which device they are using and how quickly transactions are coming in.

FD also uses graph analysis, which maps the links between accounts. That helps it spot mule rings (networks of accounts used to move stolen money) instead of judging each transaction on its own.

Every decision comes with reason codes, a plain explanation of why a transaction was flagged. Auditors and the bank’s dispute team can review them at any time.

Before switching on blocking, the bank ran FD in shadow mode on its own anonymized transaction data. The team could compare FD’s decisions with real outcomes before relying on them. The model was then tuned to the bank’s own customer patterns.

The result

After 90 days, the bank was seeing fewer account takeovers and less payment fraud, and fewer genuine customers were being blocked. The reason codes also made chargeback disputes quicker to settle. FD now covers the bank’s wallet and QR payments as well.

Questions to ask about your own fraud controls

  • How many genuine customers do our current rules block each month?
  • Could we explain every blocked transaction to an auditor or an upset customer?
  • Can we see links between accounts, or only individual transactions?