A customer who has bought from you four times files a fraud claim on the fifth order. You can win that dispute, but only if you recorded the right things during the first four.
Chargebacks get treated as a customer service problem handled after the fact, which is why brands lose disputes they should win. The determining work happens at checkout, months earlier, and it is invisible until the moment it is needed.
Two changes in 2026 made that timing gap more expensive, and the second of them creates a specific instrumentation requirement that most small brands have not acted on.
What A Chargeback Actually Is
Worth being precise, because the mechanics explain why the process feels so one-sided.
A chargeback is a forced reversal of a card transaction, initiated through the cardholder's bank rather than through you. The money moves out of your account, and you are not consulted before it happens. Chargebacks exist as a consumer protection mechanism, and the asymmetry is intentional.
Representment. The process by which a merchant contests a chargeback by submitting evidence that the transaction was legitimate and the disputed reason does not apply, effectively re-presenting the transaction to the issuing bank. It occurs after funds have already been reversed, places the burden of proof on the merchant, and is decided by the issuer rather than by a neutral party. Its success depends almost entirely on the quality of evidence gathered before the dispute existed.
The consumer protections underlying this sit in law rather than in card network rules alone, and Regulation Z is the relevant framework for credit card billing disputes in the United States. The card networks then layer their own operating rules on top, and Visa publishes its core rules publicly, which is the primary document for anything in this article.
That distinction matters practically. You are operating inside a system designed to favor the cardholder, so the objective is not fairness but evidence, and the evidence requirements are published rather than mysterious.
Friendly Fraud And Why It Grew
The category that dominates the problem for direct-to-consumer brands.
First-party misuse, commonly called friendly fraud, is a chargeback filed by a legitimate customer on a legitimate transaction. Sometimes it is deliberate, sometimes the cardholder genuinely does not recognize a charge, and sometimes it is simply easier than requesting a refund.
Vendor sources report friendly fraud accounting for up to 75 percent of all chargebacks. Treat that carefully, since it carries an "up to" qualifier and comes from companies selling tools to fight it, but the direction is consistent across the industry and matches what most operators observe.
The cardholder faces no consequences, which teaches some that they can repeat the behavior without repercussions.
The structural point is that the incentives are one-directional. A customer who files a false claim keeps the goods and the money at no cost or risk, while the merchant absorbs the loss, the fee, and the ratio damage. Nothing in the system discourages repetition.
Which explains why the networks introduced a specific mechanism for this rather than leaving it to general dispute rules, and why that mechanism is built around proving relationship history rather than proving the individual transaction.
The 2026 Threshold Change
The structural change most brands have not registered, and it has three parts.
Fraud reports and chargebacks, previously monitored under separate programs, now sit in a single ratio. Both count against the same number.
Roughly a third less room than a year earlier. A merchant comfortable under the old rule can be flagged on identical dispute volume.
The earlier program had an early warning stage lasting months. Reporting indicates that tier is gone: you are compliant or you are excessive.
Reported as a flat charge per disputed or fraudulent transaction once classified excessive, administered through the acquirer.
The combined ratio deserves particular attention because it creates a genuine unfairness. A customer who falsely claims fraud can cause a fraud report to be filed without ever pursuing a chargeback. Reporting describes the consequence directly: the merchant takes the hit to their fraud ratio and has no opportunity to prove the transaction was legitimate, because there is no dispute to contest.
That is the specific problem the evidence mechanism in section five was extended to address, and it is why a brand can see its ratio deteriorate without receiving more chargebacks.
Reporting also notes a regional exception, with merchants in one region retaining the 2.2 percent threshold, and a separate monitoring threshold for card-testing activity based on the share of authorisation attempts flagged as enumeration.
Whether It Applies To You
The scoping question most coverage skips, and the answer for a small brand is more nuanced than the headline.
The program has a minimum qualifying volume, reported at 1,500 combined fraud and dispute transactions in a month. That is disputes, not orders. A brand would need very substantial volume before 1,500 monthly disputes became plausible, so most direct-to-consumer businesses sit far below the floor.
The program also applies directly to acquirers rather than to merchants. Which is where the nuance lands.
Reporting indicates acquirers pass the pressure downstream, with stricter monitoring, higher reserve requirements and quicker terminations for merchants they consider risky. So a brand nowhere near the network threshold can still find its processor tightening terms, holding a reserve, or ending the relationship, because the acquirer is managing its own exposure.
The practical implication is that your processor's tolerance matters more to you than the network threshold does. Processors set their own limits, often well below the network's, and they are the party that can actually stop you taking payments.
So the useful question is not whether you are near 1.5 percent. It is what your processor's threshold is, what happens as you approach it, and whether a reserve or termination is the consequence. That is answerable by asking them, and very few merchants have.
Being small is not protection here. A small merchant is easier for an acquirer to terminate than a large one, since the revenue at stake is lower and the risk calculation is simpler.
The Evidence You Must Capture Now
The most important section here, and the one that turns this from background reading into a task.
Compelling Evidence 3.0 lets a merchant defeat a first-party fraud claim by proving an established relationship with the cardholder rather than by arguing about the disputed order. The qualification requirements are specific.
Read the timing requirement again, because it is the whole argument. The qualifying transactions must have occurred between four months and a year before the dispute. Whatever your checkout was recording then is what you have to work with now.
A brand that captures order details but not device identifier or IP address is relying on the other two elements, which are weaker: delivery addresses change, and guest checkout means there may be no account identifier at all.
Capture and retain device identifier and IP address on every card-not-present transaction, associated with the order, from today. Reporting suggests issuers increasingly expect device evidence and that its presence shifts the balance of proof. Nothing you do after a dispute arrives can substitute for it, so this is an instrumentation decision rather than a dispute-handling one.
The retention period follows from the rule: you need this data available for at least a year, since a dispute can reference transactions up to 365 days prior. Check that your platform is retaining rather than only capturing.
Guest checkout deserves separate thought. It removes the account identifier from the available elements and makes device and network data disproportionately important, which is a real argument for encouraging account creation beyond the marketing benefit.
Two Dates Worth Knowing
Both extensions of the same mechanism, and one of them is imminent as this publishes.
18 April 2026. Reporting indicates the framework was extended to cover non-disputed fraud reports, allowing merchants to submit historical transaction data to prevent those reports counting against their dispute ratio.
That directly addresses the unfairness described in section three. Previously a fraud report with no accompanying chargeback damaged your ratio with no route to contest it. This gives one.
24 October 2026. Reporting indicates the framework expands to multi-merchant transaction evidence for first-party fraud disputes, meaning evidence of the cardholder's transaction history need not be confined to your own business.
If this expansion arrives as reported, it materially helps brands with low repeat purchase rates, which is precisely the group least able to produce two prior undisputed transactions of their own. Worth asking your processor or dispute provider whether they support it, since the mechanism only helps if the party filing your representment uses it.
These dates come from vendor coverage rather than from network documentation retrieved directly, so verify the current position against Visa's published rules or your acquirer before relying on either. Program details in this area change, and vendors have an interest in emphasising new capabilities.
The general lesson is worth more than either date. The evidence framework is being extended repeatedly, which means the reasonable expectation is that data-quality requirements tighten rather than relax, and that brands capturing more at checkout will be better positioned each time it moves.
The Ecom Profit Box
Our collection of ecommerce growth resources, including the margin frameworks a dispute decision needs.
Get It FreeNot Capturing Device Data?
Then disputes arriving next spring are already unwinnable. We can help you work out what your checkout retains.
Book A CallPrevention Beats Representment
An economic point that is obvious once stated and routinely inverted in practice.
A dispute you prevent never enters your ratio. A dispute you win still entered it, cost you staff time, and in some arrangements cost a fee regardless of outcome. Winning is better than losing and considerably worse than not receiving the dispute.
- Clear billing descriptors. A cardholder who does not recognize a charge on a statement is the most preventable dispute there is. Your descriptor should match the name customers know you by.
- Dispute alerts. Services that notify you when a cardholder flags a transaction, giving a window to refund before it becomes a chargeback. You lose the sale and avoid the dispute.
- Accessible refunds. A customer who cannot easily get a refund from you will get one from their bank. Friction in your returns process manufactures chargebacks.
- Delivery confirmation. Removes the entire not-received category from contention when it is documented.
- Order and shipping notifications. Reduce the not-recognized category by keeping your name in front of the customer between purchase and delivery.
The refund point is worth emphasising for brands that make returns deliberately difficult to protect margin. That policy is not free: it converts refund requests into chargebacks, which cost the goods, the money, a fee, and ratio damage rather than just the refund.
Our guide to reducing return rates covers doing that through product and expectation-setting rather than through friction, which is the version that does not backfire here.
When Fighting Is Worth It
Not every dispute should be contested, and the calculation has more inputs than the order value.
The direct comparison is the recoverable amount against the cost of assembling and submitting the case. For a low-value order fought manually, the staff time frequently exceeds the recovery.
Three factors argue for fighting beyond the immediate money:
- Ratio pressure. If you are anywhere near your processor's tolerance, a won dispute removes a data point from a number that matters more than the order.
- Repeat offenders. A cardholder who has done this before will do it again. Contesting establishes that you do.
- Strong qualifying evidence. Where the relationship history clearly qualifies, the case is strong and cheap to make, which changes the arithmetic.
Against that, contest selectively where evidence is genuinely thin, where the customer has a legitimate complaint dressed as a fraud claim, and where the amount cannot justify any handling cost.
A meaningful share of disputes are customers who had a real problem and reached for the wrong mechanism. Winning those on procedure leaves an unhappy customer and an unresolved issue, and if the same product generates repeated disputes the problem is the product rather than the disputes. Read your dispute reasons as product feedback, not only as losses.
Nobody Publishes Win Rates
An absence worth reporting, since a realistic representment win rate is the number every merchant wants.
Researching this article across the dispute-management industry produced no published win rate. The closest formulation encountered was that win rates depend heavily on evidence quality and speed, which is true and is not a number.
That is notable given what these companies sell. Firms offering representment as a service, some on a performance basis, do not publish the outcome metric for the service, while publishing a great deal else including adoption statistics and threshold explainers.
The charitable reading is that the figure varies so much by category, evidence quality and reason code that a single number would be misleading. The less charitable reading is available to you too. Either way, treat any win rate quoted to you in a sales conversation as unsupported by published industry data.
Track, per month: disputes received, disputes contested, disputes won, and the staff hours spent contesting. Your win rate and your cost per win are then real numbers about your business rather than borrowed benchmarks. Most brands do not track the third and fourth of those, which is why they cannot tell whether contesting is worth doing.
That measurement also gives you the basis for evaluating a vendor. If you know your current win rate and cost per win, an outsourcing proposal becomes comparable. Without it, you are buying on an unpublished number.
The Fraud Controls That Matter
Distinguishing genuine third-party fraud from first-party misuse, since the controls differ.
True fraud, where someone else's card is used, is addressed by screening at the point of transaction: address verification, security code checks, velocity limits, and risk scoring. Those reduce fraudulent orders and, applied bluntly, also reject legitimate ones.
The tension is real and worth stating. Every tightening of fraud screening declines some good customers, and the cost of a false decline is a lost sale plus a customer who may not return. Screening tuned to eliminate fraud entirely will cost more in declined revenue than the fraud did.
Card testing is a separate problem with its own signature: large volumes of small authorisation attempts as someone validates stolen card numbers against your checkout. Reporting indicates the networks monitor this separately, based on the share of authorisation attempts flagged as enumeration. Defences are bot management and rate limiting rather than transaction screening.
The controls that serve both purposes are the ones worth prioritizing, because they generate the evidence in section five as a by-product. Device fingerprinting identifies suspicious sessions today and produces qualifying data for disputes in a year, which is unusually good value for a single implementation.
That dual purpose is the strongest argument for doing this now rather than when a dispute problem appears. By the time you need the evidence, the window to have collected it has closed.
Reading The Statistics
A short note on the figures in this space, all of which come from interested parties.
Every secondary source consulted for this article sells chargeback management, dispute automation, or fraud tooling. That does not make their explanations of the rules wrong, and the rule descriptions were consistent across sources and check against the networks' published material. It does mean the framing and the statistics need reading with the seller in view.
| Claim | How To Read It |
|---|---|
| Friendly fraud is "up to 75%" of chargebacks | An "up to" figure from vendors selling friendly fraud tools. Directionally consistent with operator experience; not a measurement. |
| Average chargeback rate rose 53% in two quarters | A large move over a short window from a vendor benchmark. Check your own trend rather than adopting the panic. |
| 90% of merchants use the evidence framework | Reported from a network survey, with another industry report giving 87%. Both describe surveyed merchants, who are not all merchants. |
| Device evidence wins "materially more" disputes | From a company selling device evidence. The mechanism is sound and consistent with the published rules; the magnitude is unquantified. |
The rules themselves are the reliable part of this topic, because they are published. Where a vendor describes a threshold, a date or an evidence requirement, that is checkable against network documentation. Where a vendor describes an outcome, it generally is not.
What To Do This Month
Ordered by what expires if you delay, which is not the same as ordered by importance.
- Confirm you capture device identifier and IP at checkout, associated with the order, and that you retain both for at least a year. Every day without this is a day of unwinnable future disputes.
- Ask your processor what their dispute threshold is and what happens as you approach it. Their tolerance governs you, not the network's.
- Check your billing descriptor against the name customers actually know. The cheapest prevention available.
- Start tracking four numbers monthly: disputes received, contested, won, and hours spent. You cannot evaluate anything without them.
- Review your refund friction. Anywhere a customer struggles to get a refund is manufacturing chargebacks.
- Ask whether your dispute provider supports the current evidence framework, including the October expansion, since the mechanism only helps if the filing party uses it.
- Read your dispute reasons as product feedback. Repeated disputes on one product describe the product.
The Point To Take Away
Chargeback outcomes are decided by instrumentation rather than by argument. The framework that defeats false fraud claims requires evidence from transactions that happened four months to a year earlier, so the work is capturing the right data now for disputes that have not been filed.
The threshold has tightened, the warning tier is gone, and pressure reaches small merchants through their processor rather than through the network. None of that is controllable. What is controllable is whether, when a dispute arrives next spring, the evidence to beat it already exists.
What To Remember
- The evidence must exist before the dispute. Qualifying requires two prior undisputed transactions from 120 to 365 days earlier, matching at least two of four data elements. It cannot be created retroactively.
- Capture device identifier and IP at checkout now, retained for at least a year, or disputes arriving next spring are already unwinnable.
- The threshold dropped to 1.5 percent on 1 April 2026, fraud reports and chargebacks now sit in one combined ratio, and the warning tier that previously gave time to correct course is reportedly gone.
- Most small brands sit below the network's volume floor and still feel it, because monitoring applies to acquirers who pass pressure downstream as reserves, tighter terms and faster terminations.
- Your processor's tolerance governs you, not the network's threshold. Ask what it is and what happens as you approach it.
- Refund friction manufactures chargebacks. A customer who cannot easily get a refund from you will get one from their bank, at far greater cost to you.
- No vendor in this industry publishes a representment win rate, so track your own disputes received, contested, won and hours spent rather than accepting a benchmark.
Where This Came From
- Visa's publicly published core rules, the primary document for dispute conditions, evidence requirements and monitoring programs referenced throughout this article.
- Consumer Financial Protection Bureau, Regulation Z, the framework underlying credit card billing dispute rights in the United States on which card network rules are layered.
- Payments industry reporting on the merchant monitoring threshold falling from 2.2 percent to 1.5 percent effective 1 April 2026, the combination of fraud reports and dispute transactions into a single ratio, the removal of the previous warning tier, a reported flat fee of $8 per disputed or fraudulent transaction once classified excessive, a minimum qualifying volume of 1,500 combined dispute transactions monthly, a regional exception retaining the 2.2 percent threshold, and a separate monitoring threshold for card enumeration activity.
- Payments industry reporting on Compelling Evidence 3.0 qualification, requiring two or more prior undisputed transactions occurring 120 to 365 days before the disputed transaction and matching at least two of four data elements comprising customer account or login identifier, delivery address, device identifier or fingerprint, and IP address, with liability shifting to the issuer where the test is met.
- Payments industry reporting on framework extensions, indicating coverage of non-disputed fraud reports from 18 April 2026 and expansion to multi-merchant transaction evidence for first-party misuse disputes from 24 October 2026. These dates originate in vendor coverage rather than network documentation retrieved directly, and readers are advised in section 06 to verify the current position with their acquirer.
- Every secondary source consulted for this article sells chargeback management, dispute automation or fraud tooling, as stated in section 11. Statistics cited include friendly fraud representing "up to" 75 percent of chargebacks, an average chargeback rate of 0.26 percent in Q3 2025 reported as a 53 percent increase over Q1 2025, and framework adoption reported at both 90 percent from a network survey and 87 percent from an industry report. No source consulted published a representment win rate, an absence discussed in section 09.