The chargeback signal problem: why disputes and fraud aren't the same thing

Thomas Dodd, Consultant at KAE, on why disputes and fraud aren't the same thing, and how merchants can build fraud strategies that look beyond the chargeback.

A customer receives a notification for a transaction they don’t recognise; they log into their bank account and find their card made an online purchase whilst they were at work. The customer is a victim of fraud, perhaps their card credentials were stolen, so they dispute the transaction. A chargeback is initiated and the customer’s card issuer reverses the payment, moving the funds back to the customer’s account. In this quintessential case of stolen credential fraud, the chargeback flags that fraudulent payment activity has taken place. Chargeback data has often been used as a signifier of payment fraud and enabled issuers and merchant acquirers to recognise and try to prevent fraudsters. 

Yet, as the exponential growth of global ecommerce activity has led to the evolution of payment fraud into a fully-fledged industry, chargebacks alone are no longer reliable indicators of fraud and, often, can be misused as a mechanism of fraud. As losses from fraud ballooned to over USD 579bn in 2025, measuring and managing fraud is more important than ever.

This article breaks down how chargebacks have been used previously, why they are no longer a robust fraud indicator, the potential costs for businesses who lag, and how fraud can better be managed to meet emerging challenges.

Chargebacks: a dated tool for fraud identification

Chargebacks are a card scheme mandated mechanism that enables customers to claim back their money from a transaction if there is a problem with their purchase, such as if the transaction is fraudulent. 

When a customer initiates a chargeback, the card issuer reviews the dispute and, if the dispute is valid, the issuer reverses the initial transaction and funds are removed from the merchant’s account to the cardholder’s account. The chargeback captures the customer’s reason for the dispute through reason codes, and although each network tends to have slightly different codes, fraud reason codes can sometimes identify the transaction environment and the volume of transactions. Therefore, chargeback data has acted as a useful proxy enabling card issuers and processors to assess levels of fraudulent activity and to identify common fraud environments.

While chargebacks have historically provided crucial context to fraudulent transactions, the rapid development of payments fraud has rendered them an oblique fraud signal as different causes of fraud can produce identical chargeback records. Two fraud types illustrate this: 

1. Account takeover (ATO) fraud: fraudsters gain access to a customer account, for example an ecommerce account, using phishing, malware, or data breaches. They then use stolen credentials or stored payment credentials to make unauthorised purchases on the customer’s account.
2. First-party fraud: involves no external fraudster; first-party, also known as friendly fraud, occurs when a customer deliberately or accidently disputes a legitimate transaction. According to LexisNexis, first party fraud accounted for almost 40% of fraud cases amongst their clients in 2025.

Despite having different root causes each of these two fraud categories can produce identical chargeback reports. When the customer reports a chargeback and the original transaction is flagged as fraudulent, these two fraud types could produce reports that flag a CNP transaction that was not authorised by the customer. Beyond that, a chargeback report offers limited insight outside of payment losses. In the case of first-party fraud, the chargeback itself is a vehicle for fraud, muddying the water as to what is a real case of fraud versus a false customer dispute.

Evolving fraud

Increasingly, fraud is being moved to new frontiers that entirely remove chargebacks from the equation. Synthetic identity fraud is an emerging payment fraud that involves fraudsters creating fake digital identities that are manufactured using a combination of real personal identifiers and fictious names, dates, and addresses. Since the identities created are false, there is no real customer to submit a chargeback on the fraudulent transaction. Instead, either the fraud is spotted or it goes undetected. 

Synthetic identity fraud has reportedly been supercharged by generative AI, as fraudsters leverage it to create increasingly complex profiles that mirror the activity of a real account. The deployment of AI at scale can be seen in the figures. LexisNexis reported that, in 2024, synthetic identity fraud accounted for 1.4% of cases amongst their clients; in twelve months this figure jumped to 11%. 

The stakes

For businesses around the world, it is imperative to be aware of the evolution of fraud and the limitations of relying on signals that do not offer granular insight on fraud techniques and causes. The financial implications of fraud can be devastating, Transunion surveyed US businesses in 2025 and found that, on average, companies lost close to 10% of their revenue due to fraud over a twelve-month period. More concerningly, this represented a 46% increase on the year prior, underlining the industrial scale growth of fraud fuelled by widespread access to sophisticated AI models. The numbers highlight that fraud is becoming harder to trace and increasingly difficult to prevent, and the cost of inaction for businesses is only likely to get higher.

Meeting current and future fraud threats

The scale and complexity of payment fraud mean that there is no catch-all signal that accurately measures fraud. Instead, companies need to adopt a multi-layered approach to preventing fraud, including leveraging dispute ratio data. 

First party fraud can be partially mitigated against by ensuring that refunds are granted to consumers who complete the refund process. Chargebacks typically incur high fees from payment acquirers; it is often best to avoid these by simply granting refunds. However, during the refund process, merchants can add complexity to discourage consumers from casual fraud. For example, returns could require customers to login to an account, fill out specific details of why the product is being returned, or require physical drop-off in-store. By following these two steps, costly chargebacks can be reduced, and the volume of returns can also be steered downwards.

Meeting genuine fraud threats from ATO and synthetic identity fraud is challenging and requires leveraging data that adds context to customers. For instance, where a merchant can identify the usual IP address of a customer in one location, an ATO attempt can be stifled if a transaction is attempted from a different jurisdiction. With synthetic identity fraud, unusual behaviour can be even more difficult to flag since the identity is fictitious and not comparable to the behaviour of a legitimate customer. Therefore, specialised companies ensure the profile can be verified against trustworthy data points, such as cross-referencing personally identifiable information, checking for government-issued identification, and even testing if the profile can follow live camera instructions. 

Chargeback dispute data can initially be used in conjunction with these approaches; where a merchant sees high rates of chargeback transactions, it can be seen as an identifier that the merchant is exposed to fraud and needs to act. However, in order to identify the cause of the fraud and the subsequent steps to mitigating against further fraud losses, chargeback reports alone are not sufficient.

Building a fraud strategy beyond the chargeback

Fraud has developed faster than the tools traditionally used to measure it. Chargebacks remain a useful data point, but relying on them as a standalone indicator leaves gaps as some fraud types produce no dispute at all, and others produce a dispute for reasons unrelated to the underlying fraud.

The practical question for issuers, acquirers, and merchants is which additional signals to add, and how to weight them for their own customer base. This varies by market, product, and competitor. What counts as unusual IP or device behaviour, how much friction a refund process can introduce before it affects genuine customers, and where synthetic identities are most likely to appear all depend on the specifics of a business's customers, competitors and market. For businesses reviewing their fraud strategy, the starting point is a clear picture of where fraud is occurring and why chargeback data alone no longer explains it.

About the author

Tom is a Consultant at KAE, specialising in payments and financial services strategy. He has delivered engagements spanning product refinement, cross-border payments, and market strategy, bringing a detail-oriented approach to translating complex market dynamics into actionable insight for clients.

 

 

About KAE

KAE helps banks, payment companies, and fintechs around the world make data-backed marketing, product, and strategy decisions by uncovering deep customer, market, and competitor insights. For over 30 years they have provided market-leading businesses with the customer and market intelligence needed to shape and execute go-to-market strategies, develop products and experiences, optimise pricing and communications, and build effective sales and partnership ecosystems.