Friendly Fraud Isn’t Very Friendly

By Matthew Goldman, founder of Totavi and self-proclaimed 20-year payments nerd. Originally published on CardsFTW.

I have always thought “friendly fraud” was a strange term. Plenty of legitimate reasons exist for someone to dispute a transaction they actually made: merchant descriptors remain terrible, families share cards, subscriptions get forgotten, and sometimes a transaction from three weeks ago simply doesn’t look familiar when it shows up on a statement. I have written before about how bad transaction data contributes to this problem, and those cases are more customer confusion than fraud. However, another category exists where a cardholder buys something, receives it, knows perfectly well they bought it, and then tells their bank they didn’t. Calling that “friendly fraud” feels a little like calling shoplifting “friendly inventory shrinkage.”

The industry increasingly uses “first-party fraud” for the broader category, which I prefer because it is more descriptive: the person committing the fraud is actually the customer. 

Socure defines first-party fraud as a consumer using their own legitimate identity to commit a dishonest act for financial gain, and their consumer research suggests this is not a tiny edge case. Thirty-five percent of respondents admitted to at least one form of first-party fraud, including 22% who said they had requested a refund for an online purchase they actually received and 20% who said they had disputed a legitimate financial transaction. More than half of Gen Z respondents said they would commit first-party fraud if they knew there would be no negative consequences. (I am somewhat skeptical of surveys where people self-report committing fraud, but the direction is hard to argue with.) 

The Internet might think “friendly fraud” is a life hack. I think it is criminal. It also causes major problems for issuers, merchants, and credit card companies.

Disputes Have Gotten Much Easier

I remember when disputing a transaction generally meant calling the number on the back of your card, navigating a phone tree, waiting for an agent, explaining what happened, and then potentially filling out paperwork. Nobody wants to go back to that experience. One real benefit of modern digital card servicing is that I can open an app, tap a transaction, answer a handful of questions, and initiate the dispute without ever talking to another human being. The industry has spent years making it easier for legitimate cardholders to dispute transactions, which is generally a very good thing, while also creating a system that is easier for people to abuse. 

Consumers clearly value this: Javelin’s 2026 chargeback study found that 38% of consumers disputed at least one debit or credit transaction in the prior 12 months, while 63% said the ability to easily dispute erroneous or fraudulent charges was an important consideration when choosing a new credit card. Ease of disputes actually ranked above rewards-related features, status, card material, and affinity with a particular brand.  

The problem, of course, is that removing friction from something generally causes people to do more of it. Javelin found that 83% of issuers attribute rising dispute activity to greater consumer awareness of the process. In the same study, 49% of issuers cited increasing first-party fraud and 25% specifically pointed to easier dispute channels, such as initiating a dispute directly from mobile banking. 

This puts issuers in a somewhat awkward position because making disputes harder would solve the wrong problem. If my physical card is stolen and someone spends $800 at Best Buy, I should be able to tell my issuer about it quickly. If a merchant charges me twice, I should not need to find a fax machine. The industry was right to make these processes easier, especially since the dispute experience is one of those moments where a cardholder’s opinion of an issuer can change very quickly.

Javelin found that issuers frequently saw card usage increase after a dispute was resolved in the consumer’s favor and decrease when it wasn’t. This makes intuitive sense. Rewards might get somebody to apply for a card, but servicing is what determines whether they still like the card after something goes wrong.

The challenge is that the customer experience on the front end has modernized much faster than the machinery sitting behind it.

The Backend is Not As Modern (and it is Expensive)

A modern card program can authorize transactions in milliseconds, provision cards instantly into Apple Pay, run fraud models in real time, dynamically control spend, and integrate with dozens of third-party services through APIs. Then someone disputes a $47 transaction and the process can suddenly involve case queues, network rules, reason codes, screenshots, receipts, merchant evidence, filing windows, provisional credits, and people manually deciding what to do next.

Visa and Mastercard have extensive rules on what qualifies for a chargeback, when something must be filed, which party is liable, what evidence is required, and how a case progresses. Regulatory requirements also govern how issuers handle certain consumer disputes. The problem is that much of the operational model grew up around the assumption that humans would navigate this complexity case by case.

That works reasonably well until the program grows, because not every dispute is worth spending much money investigating. If somebody disputes a $22 transaction and it costs $30 in labor and operational overhead to determine whether they should get the $22 back, there is an obvious temptation to accept the dispute. That may be the rational economic decision in an individual case, but doing it repeatedly teaches people that disputes are essentially free money.

The opposite approach is not particularly attractive either. An issuer can build stricter rules and reject more disputes, although that risks denying legitimate claims and damaging the customer relationship. A traditional rules-based system also struggles with all the nuance involved in determining whether a particular dispute is suspicious. Someone disputing three legitimate-looking transactions across three years is very different from someone who has disputed 18 transactions across multiple merchants in six months, even if the transaction in front of the analyst looks identical.

The underlying fraud losses get most of the attention, although dispute-related servicing costs are substantial on their own. Socure’s 2023 report estimates that between 50 million and 100 million disputes occur annually, at a processing cost of $10 to $50 each. The merchant side looks similarly painful. Javelin found that merchants reported an average of $82 in internal costs plus $46 in third-party fees for each chargeback they handled. Again, those are merchant costs, not issuer costs. Still, the absurdity is the same: The industry can easily spend more arguing about a relatively small transaction than the transaction itself was worth.

There is also a compounding problem here. Fraud losses increase as transaction volume grows, while operational costs increase because more transactions create more disputes and more disputes create more work. Unlike authorization processing, where another million transactions mostly means more compute, a manually intensive disputes organization frequently needs more people as volume rises.

This is the more interesting part of the dispute problem. First-party fraud is expensive, certainly, but the system we’ve built to decide which disputes are legitimate can itself become enormously expensive. If an issuer responds by simply approving more claims because investigation costs too much, the operational problem then contributes directly to the fraud problem.

So issuers wind up stuck between three bad choices: spend a lot of money investigating low-dollar disputes, accept a lot of claims because investigating them costs too much, or tighten the process in ways that make legitimate customers miserable.

Not an ideal operating model.

The Answer Isn’t More Dispute Analysts

This is where Decisionly comes in. The founders previously built Chargehound, an automated chargeback product for merchants that PayPal acquired, and Decisionly essentially attacks the same class of problem from the issuer side. The premise is that the dispute process should be designed around automated investigation and decisioning, with humans handling exceptions, rather than assuming every dispute needs to become a task for an operations employee. 

A lot can be evaluated programmatically before anyone touches the case. Decisionly checks things like EMV and 3DS liability shift, filing deadlines, network-required chargeback conditions, reason codes, authentication data, and patterns associated with first-party fraud. It also uses AI to review unstructured material such as receipts, images, case notes, and other supporting documentation. 

That distinction matters because “AI for disputes” can otherwise sound like another application where somebody added a chatbot to an existing workflow. The more useful idea is that most disputes should never enter a manual workflow in the first place.

Bilt is one example. As its dispute volume grew, Decisionly says Bilt needed to increase capacity without increasing headcount. After implementing the platform, Bilt eliminated 80% of manual dispute work, reduced resolution time by 20%, and processed exceptions twice as quickly. Tallied, a card program manager recently acquired by FinWise Bank, reports automating 96% of disputes with a 93% win rate and no incremental operational lift for each additional program. These are customer-reported Decisionly results rather than benchmarks every issuer should expect to reproduce, although they demonstrate what a very different dispute operating model can look like.

I especially like the implications for smaller issuers and fintechs. One persistent challenge in launching a card program is that consumers implicitly compare the experience against Chase, Capital One, or Amex, even though the new issuer obviously does not have a 50,000-person servicing organization. Much modern fintech infrastructure has focused on closing that gap by turning capabilities that once required massive internal teams into software services. Disputes seem like a natural next place for that transition.

Make the Back End Smarter, Not the Front End Harder

The easy conclusion to draw from the growth of friendly fraud is that people are abusing disputes, so we should add more friction. I don’t think the industry should end up there. (In essence, be less friendly to your users, the good ones, too).

Consumers should be able to dispute a transaction from their phones. Legitimate fraud victims should get fast resolutions, and nobody should have to sit on hold for an hour because a merchant charged them twice. The fact that some people abuse a good customer experience is not a strong argument for making the experience bad for everyone else.

The better answer is to make the system behind the button substantially smarter. A dispute from someone with years of normal behavior should not necessarily be treated the same way as the 15th dispute from someone exhibiting obvious patterns of first-party fraud. A $12 dispute that clearly satisfies network requirements should not require the same human effort as a complicated $2,000 case with conflicting evidence. Merchant documentation that clearly does not support the transaction should not sit in a queue waiting for someone to read it three days later.

We have already rebuilt most of the modern card stack around this idea. Computers handle the straightforward decisions at enormous scale, while humans spend their time on the difficult cases where judgment actually matters. Authorization works this way. Fraud detection works this way. Increasingly, underwriting works this way.

Disputes should too.

Let’s reserve the word “friendly” for the actual customer experience, rather than the fraud.

CardsFTW

CardsFTW is authored and published by Matthew Goldman and the team at Totavi, a boutique consulting firm specializing in fintech product management & marketing.