Every dispute team has a version of this story: a cardholder who files claim after claim, always just plausible enough to pass on its own. When reviewed one at a time, none of them looks unusual. When reviewed together, the pattern is obvious.
That gap between "unusual in isolation" and "obvious in aggregate" is where serial disputers operate, and it's expensive. Each unchallenged claim doesn't just cost a refund. It signals that the system doesn't connect the dots, inviting the abuser to keep going.
Why is this harder than it looks?
The difficulty isn't spotting one bad claim; dispute teams are generally good at that. The difficulty is that repeated abuse is built from individually unremarkable actions spread over time and, often, across merchants. A fraud reason applied to a transaction that the cardholder authorised. A new dispute opened within days of the last one being rejected. None of these trip an alarm on their own, and a case-by-case review process was never designed to hold all of them in view at once.
The cost of missing it compounds in three directions:
- Merchants absorb chargebacks for transactions that were never actually fraudulent, which erodes the relationships banks depend on.
- Dispute teams devote equal effort to every claim, regardless of risk, so the highest-risk cases receive the same scrutiny as the most routine ones.
- Genuine victims of fraud wait longer for resolution because the queue isn't prioritised by anything more sophisticated than order of arrival.
None of this is about unfairly labelling cardholders. The point of getting better at detection isn't suspicion for its own sake, but making sure the small number of cases that warrant a closer look actually get one, so the rest can move through quickly and fairly.
The scale of the problem
This isn't a marginal issue. Friendly fraud, disputes filed against transactions that were in fact legitimate, is now estimated to cost businesses in the US over $100 billion a year and accounts for roughly three-quarters of all chargebacks industry-wide, a pattern that has become increasingly common alongside the shift toward e-commerce and away from cash. Some of that is genuine confusion. But repetition is where it stops looking accidental.
The trend is accelerating rather than levelling off. First-party fraud now accounts for over a third of all global fraud cases, having more than doubled its share in a single year, and more than 64% of enterprise merchants report that friendly fraud has increased over the past three years, with one-quarter reporting increases of 25% or more.
Each dispute also carries a direct processing cost for the issuer, estimated by Mastercard at roughly $9-10 per case regardless of outcome, before accounting for customer relationship risk: Javelin Strategy & Research suggests 60-70% of customers who feel a dispute was handled poorly will move to another bank entirely.
None of this is about any one institution falling behind. It reflects a category of fraud that has grown faster than most legacy dispute processes were built to handle, which is exactly why better detection has become a shared priority across the industry rather than a niche concern.
What good detection actually requires
Getting this right means treating every claim as part of a history rather than an isolated event. That means holding a cardholder's full record in view when a new dispute lands: how many claims they've filed, how those claims were resolved, and whether the behaviour around the claim, its timing, its consistency, its pattern after a rejection, adds up to something worth a second look.
Assembling that picture by hand, case by case, is exactly where dispute teams lose time and consistency. It requires pulling history from multiple systems, remembering which merchants a cardholder has previously disputed, and making judgment calls that vary from one agent to the next. It's precisely the kind of task that benefits from being handled automatically and consistently, so the human decision at the end is better informed rather than replaced.
How Amiko brings this into the workflow
This is the problem Amiko's serial disputer intelligence is built to address. Rather than asking agents to manually cross-reference a cardholder's dispute history, Amiko surfaces it automatically the moment a case is opened: prior claims, how they were resolved, and why the current case has been flagged for closer attention.
What makes this more than simple pattern-matching is that Amiko reasons across the full picture the way an experienced investigator would, recognising when a cluster of individually minor details adds up to something worth escalating, and when a higher claim count is genuinely just bad luck. The result is a judgement an agent can trust and act on immediately, not a rule that fires without context.
That same intelligence is also starting to move earlier in the journey. Rather than informing the agent only after a case has already been filed, Amiko is extending this evaluation into the intake conversation itself, so that a cardholder's history can shape the experience from the moment they begin reporting a dispute, before the case ever reaches a queue.
The effect is a shift from reacting to repeat abuse after the fact to catching it at the point of submission, without adding friction for the cardholders who don't need it.
Serial disputer behaviour will continue to evolve, and no single rule can fully capture it. What matters is building the capability to see the pattern and to act on it with better information than a single claim can ever provide.
For a fuller picture on why serial disputers and refund fraud-as-a-service are becoming a bigger problem for issuers, we go deeper in a companion piece.