TL;DR
- Once a chargeback is filed, the two remaining paths, fighting and keeping, end in the same permanent mark on your ratio. They differ only in cash and time.
- Merchants win about half of the disputes they choose to fight, but across all disputes, only about a tenth of the disputed money ever comes back. The win rate is real; the money mostly is not.
- The only decision that erases the mark happens before the chargeback is filed, in the 24 to 72 hour alert window.
- For a monthly subscriber, a fourth option usually beats all three: refund the charge and keep the customer.
Who this helps: you have a dispute queue, a vendor telling you to fight everything in it, and a feeling that the math is being done for you. Here is the math with nothing attached to it.
Most advice about fighting chargebacks comes from companies that are paid a percentage of whatever gets recovered. That does not make the advice wrong, but it makes it worth checking. So check it: merchants win about half of the disputes they fight, and recover about a tenth of the money at stake. The distance between those two numbers is what this post explains.
What are the three decisions?
Every dispute ends in one of three ways, and each spends a different resource. A refund spends money to prevent a mark: you return the charge voluntarily, before a chargeback exists, and nothing enters your count. A fight spends time to recover money: the chargeback has already happened, the funds are already gone, and you submit evidence to pull them back. Keeping spends nothing: you accept the chargeback, keep the loss where it fell, and keep your team's hours.
The order matters because the options expire. The refund is only available before the bank files, which in practice means during a support conversation or inside the 24 to 72 hour window a chargeback alert gives you. Once the chargeback lands, only fight and keep remain, and both of them carry the mark.
Why doesn't winning a dispute clear my record?
Because the card networks count the dispute, not the outcome. A chargeback enters Visa's VAMP ratio and Mastercard's ECP count the day it is filed. Win the case three weeks later and the money comes back; the mark does not. Your processor and acquirer watch the same filed-not-won number. That is why fighting is a cash decision, never a ratio decision, and why a merchant near the 1.5% line gets nothing from a winning streak.
A win is also less final than it sounds. The issuer can reopen a case you just won through pre-arbitration, sometimes called a second chargeback, usually claiming new evidence. Decline that, and the case can go to arbitration, where the network itself rules and the loser pays the costs: roughly $500 at Visa and $400 at Mastercard. On a $29 monthly charge, losing that round costs more than a dozen of the sales you were fighting over.
What is the real win rate when you fight?
About half of the fights, and about a tenth of the money. US merchants win 54% of the disputes they take on, by Mastercard's survey, and 43.8% by the 2026 Chargebacks911 field report. But merchants only fight a minority of their chargebacks, some wins are reopened in pre-arbitration, and some wins come back as new disputes from the same customer. Stack those leaks and the same field report finds that only 10.7% of all disputed dollars actually come back, a number it calls net recovery, with merchants winning 8.1% of all disputes outright.
None of this means never fight. It means the headline number a success-fee vendor shows you describes their best case, not your expected one. Plan with the money number, not the win number, and the fights worth picking become obvious.
When is fighting worth it?
When the expected recovery clears the real cost of fighting, which happens far above the price of a typical monthly charge. The arithmetic has three parts: the amount times your real chance of winning, minus the time (the 2026 Mastercard/Javelin study puts average internal handling at $82 per chargeback), minus the fee risk (at Stripe, countering costs $15 that returns only if you win; many mid-market acquirers charge more and return nothing).
Run it on a $29 monthly charge with ordinary evidence: you win half of these, so the average fight brings back about $15, and the fee risk plus the first twenty minutes of anyone's time already spend it. The same math on a $290 annual plan with login records and usage history clears easily. Where your floor sits depends on what a fight costs you. With a manual, from-scratch evidence process, the $82 average is your cost, and the floor lands near $150. With templated evidence and an automated flow, handling drops far enough that a $50 to $100 floor holds. Either way, disputes below your floor are refund-or-keep decisions, not fights.
Two programs move the odds for one specific claim. When a real customer says "I did not make this purchase," Visa's Compelling Evidence 3.0 and Mastercard's First-Party Trust program let you answer with history: show the same customer made two earlier purchases they never disputed, and the loss moves back to the bank. A subscriber on month six has that history by definition. For that claim, on those cards, fighting is closer to paperwork than a gamble.
What should a subscription business do differently?
Add the branch the framework is missing: refund and keep the subscriber. A $29 dispute from a customer who stays subscribed eight more months is a $232 relationship arguing with you about $29. Fighting brings back $15 on average, burns the relationship, and the mark stays either way. Refunding at the alert stage costs the $29 plus the alert fee, erases the mark, and, handled with an honest message, often keeps the $232. The intelligence layer exists to make that comparison per customer instead of per policy: what they are worth, what they have disputed before, and which cycle the trouble comes from.
And move the decision earlier, which is what the prevention stack exists for. Every option this post prices gets better before the chargeback exists: the refund that prevents the mark is only available in the alert window, and the dispute that never files costs the $29 alert fee instead of $128 in fees and handling. The framework is not really refund, fight, or keep. It is decide early, and refund, fight, or keep is what deciding late looks like.
A chargeback counts against your ratio the day it is filed, win or lose. Merchants win about half of the disputes they fight but recover about a tenth of the disputed money. The only decision that erases the mark is the refund made before filing, and for a subscription business that window is the alert.
FAQ
Should I fight every chargeback?
No. Fighting is a cash decision that never repairs your ratio, and the real cost of a fight (about $82 of handling time on average, plus fee risk) exceeds the expected recovery on small transactions. Set a floor, fight above it with strong evidence, and resolve disputes below it at the alert stage or accept them.
What percentage of chargebacks do merchants actually win?
Two honest numbers: US merchants win 54% of the disputes they choose to fight (Mastercard survey; about 43.8% in the 2026 Chargebacks911 field report), but only about 10.7% of all disputed dollars actually come back, because most chargebacks are never fought and some wins are reopened. Plan with the second number.
Is it better to refund than to let a chargeback happen?
Almost always, if the refund comes first. A refund costs the sale and nothing else; a chargeback costs the sale, a dispute fee, handling time, and a permanent mark on the ratio Visa and Mastercard monitor. A refund issued inside the alert window prevents the chargeback from existing at all.
What is pre-arbitration on a chargeback?
The issuer's option to reopen a dispute the merchant just won, usually claiming new evidence; it is sometimes called a second chargeback. If neither side yields, the case can go to arbitration, where the card network rules and the loser pays roughly $500 at Visa or $400 at Mastercard. This reopening risk is one reason net recovery runs far below the headline win rate.
When is fighting a chargeback worth it?
When the amount times your realistic win odds clears the cost of fighting: about $82 of average handling time plus your processor's fee terms. In practice that means high tickets and annual plans with strong evidence, especially "I did not make this purchase" claims where Compelling Evidence 3.0 or First-Party Trust history qualifies. A typical monthly charge sits below the line at any handling cost.
Facts checked against Mastercard's published survey of dispute outcomes, the 2026 Chargebacks911 Field Report, and the 2026 Mastercard/Javelin cost study. Last reviewed September 8, 2026.
