The dispute that wasn't abuse
Serial chargebacks scream 'friendly fraud' — but sometimes the customer is genuinely the victim, and auto-fighting them costs you a loyal one. This week, the dispute is legit.
What we're seeing
Friendly fraud' — a cardholder disputing a charge they actually made — is real and costly, and issuers increasingly auto-score repeat disputers as abusers. But the same surface pattern (several disputes) also describes a genuine victim of card theft or a string of merchant errors.
Treat every frequent disputer as an abuser and you punish exactly the customers who were defrauded — and you teach them to bank somewhere else.
Why your current stack misses it
- A count-based rule sees 'N disputes in M months' and flags abuse — it can't tell whether the disputes are corroborated by real fraud signals.
- Context (a confirmed account compromise, consistent merchant-error stories, a long clean history) is invisible to a threshold.
The signal pattern
- Disputes line up with an independently confirmed card-compromise or breach window.
- Disputed merchants and amounts are inconsistent with the customer's own spending fingerprint.
- A long prior history with zero disputes, then a sudden cluster — an event, not a habit.
- The customer cooperates: files reports, answers questions, behaves like a victim rather than a fraudster.
What you'd do Monday morning
- Score disputes against corroborating fraud signals, not just a raw count.
- Separate 'victim of theft' and 'merchant error' from true first-party abuse before auto-representment.
- Protect long-tenured, cooperative customers from being mislabeled — measure the retention cost of false 'abuser' flags.
Spot the Fraud
Read the case. Make the call. See how you score against The PreCogs.
A cardholder has filed four disputes in two months — your abuse model wants to brand them a serial chargeback abuser. Look closer before you write them off. Clear it (genuine), or hold it (abuse)?