How to fight a chargeback, and when to let it go

Find the response deadline before you do anything else. It's in the dispute notice from your processor, usually somewhere between 20 and 30 days out depending on the card network and the reason code, and it's not negotiable. Miss it and you lose automatically, no matter how good your evidence was. Put the date in a shared calendar the same hour the notice lands.

That's the whole emergency. Everything after this is a business decision.

What a chargeback actually is

It's not a refund request. The cardholder called their issuing bank, the bank pulled the money out of your account, and now you're being invited to prove you shouldn't have lost it. The customer's bank decides. You're arguing to a third party who has never met you and is working from a reason code.

You'll also get hit with a dispute fee from your processor, commonly somewhere in the fifteen to twenty-five dollar range in the US. Check your merchant agreement for the exact number and for whether you get it back if you win, because that varies by acquirer and a lot of merchants assume wrongly.

The part people miss: winning doesn't delete the chargeback from your ratio. Card network monitoring programs count disputes received, not disputes lost. The commonly quoted trigger points are around 0.9 percent of monthly transactions for Visa and about 1.5 percent for Mastercard, with a minimum count in the hundred range. Confirm the current numbers with your acquirer, since the networks revise them. But understand the implication. If you're near the line, fighting harder doesn't save you. Preventing the dispute does.

The arithmetic that tells you whether to fight

Work out your real cost to contest one. Say it takes a staff member 25 minutes to pull the order record, the delivery confirmation, the terms the customer agreed to, and write the rebuttal. At a loaded cost of thirty dollars an hour, that's about twelve or thirteen dollars of labor. Add the dispute fee if you don't get it back.

Now weigh that against the transaction value and your honest win rate on that reason code. Not your hoped-for rate. Go back through six months of outcomes and count.

If you win roughly half of "product not received" disputes and your average order is forty dollars, you're recovering about twenty dollars per attempt against thirteen dollars of cost. Thin. At a hundred and eighty dollars an order, it's obvious. Set a floor in dollars and stop agonizing case by case.

One exception to the math. Fight a fraud-coded dispute even when it's small if the same card, device, or shipping address shows up more than once. You're not fighting for forty dollars. You're building a record.

The disputes worth fighting

Product not received, with tracking. Delivery confirmation to the billing zip code, ideally with signature on higher-value orders. This is the cleanest win in the business. If you ship to an address that doesn't match billing, your odds drop a lot.

Not as described, when you have the listing. Pull the exact product page as it appeared on the order date, the specification the customer saw, and any photos. If your descriptions are vague, you'll lose, and that's a merchandising problem rather than a disputes problem.

Subscription cancellation claims, when you have the log. Timestamped record of what they agreed to, the renewal notice you sent, and no cancellation request on file. Win rate on these is strong if your system actually logs the consent. If it doesn't, fix that this quarter.

Repeat-customer fraud claims. Visa's compelling evidence rules let you submit prior undisputed transactions from the same cardholder, tied together by matching data points like device ID, IP address, or shipping address, within a defined age window. Ask your processor whether you're eligible and what format they need. It has turned a lot of losing fraud disputes into wins for merchants selling to repeat buyers.

The disputes to let go

Digital goods with no login record or IP capture. You have nothing to show.

Any dispute where you genuinely owed a refund and didn't issue it. Fight that and you're spending money to look bad.

Card-not-present fraud where the AVS and CVV checks came back mismatched and you approved the order anyway. The evidence works against you, and you're funding somebody's time to lose.

Small-dollar disputes from a customer with a long, clean history. Eat it. Call them. Half the time it's a family member who didn't recognize the charge, and you keep an account worth more than the order.

And if you're inside a monitoring program, let the marginal ones go entirely and put those hours into prevention. That's counterintuitive and it's correct.

Writing the rebuttal so a stranger can read it

Answer the reason code. Not your version of events. If the code is "credit not processed," the reviewer wants to know whether a credit was owed and whether you issued one. A paragraph about how difficult the customer was does nothing.

One PDF. Two to five pages. Page one is a plain cover summary in short sentences: what was sold, when it shipped, who signed for it, what the customer agreed to, and why the claim doesn't hold. Then the exhibits behind it, labeled and in order.

Redact nothing that matters and include nothing that doesn't. Reviewers spend minutes, not hours. Screenshots should be legible at normal size, and every date in your narrative should appear in an exhibit.

Don't refund a customer after a chargeback has been filed. You'll pay twice. If you want to make it right, wait for the outcome, or withdraw the dispute through your processor first.

The part that actually moves the number

Your billing descriptor. If it reads as a holding company name nobody has heard of, you're generating disputes from your own happy customers. Put a recognizable trade name and a working phone number in it. This one change has cut dispute volume for more merchants than any evidence strategy ever has.

After that, look at alerts. Verifi and Ethoca both run networks that notify merchants of a dispute before it becomes a chargeback, and some programs will auto-refund qualifying transactions so the chargeback never files. You pay per alert. If your average order is small and your dispute volume is meaningful, run the numbers, because the alert fee is usually cheaper than the dispute fee plus the lost goods.

Then read your own data monthly by reason code. If one SKU, one ad creative, or one shipping lane produces most of your disputes, you don't have a chargeback problem. You have a product page that overpromises or a courier that leaves boxes on the sidewalk.

If you lose

You can usually push to pre-arbitration and then arbitration. Read the fee schedule before you do. Network arbitration filing fees run into the hundreds of dollars and you forfeit them if the ruling goes against you. For a two-hundred-dollar order, that's a bad trade almost every time. For a recurring contract worth thousands, or where the same buyer has done it three times, talk to your acquirer about whether the case is strong enough to be worth the filing.

Most of the time, take the loss, write down what the customer's bank found persuasive, and go change the thing that let it happen.

Pull last quarter's disputes tonight and sort them by reason code. Whatever tops that list is next quarter's project.