The Campaign That Killed the Merchant Account

You had the best week the store has ever had. Then 6 weeks later the payouts stopped, risk asked for shipping records, and a month after that the acquirer closed the account. Nothing about the business changed except the volume. This is the exact mechanism that turns a good month into a terminated merchant account, the numbers the card networks actually use, and what still moves them inside the window you have left.

The spike doesn't kill you. The month you pull back does.

The dispute ratio that gets you terminated is a fraction, and almost nobody watching it thinks hard about the bottom half. Mastercard's Excessive Chargeback Program divides the chargebacks received in a given month by the count of sales processed in the month before. June chargebacks over May sales. Visa's monitoring program is same month: fraud reports plus disputes over settled card not present transactions in that period. Either way the denominator is a month of sales, and disputes do not arrive in the month the orders were placed.

So run the arithmetic. Mastercard volume only, because that is what its program measures.

Say 1.3% of each cohort eventually disputes, which is not a catastrophic rate for cold traffic on a physical product. The month 2 cohort is 234 chargebacks, and they do not land together. Call it 30% in the month after purchase and 70% the month after that: about 70 show up in month 3 and about 164 in month 4.

Month 3 looks survivable. 70 from the spike plus 45 still landing from month 1's orders is 115 chargebacks, measured against month 2's 18,000 sales. That is 0.64%. The count is over 100, the ratio is nowhere near the line, and nobody escalates.

Month 4 ends it. 164 from the spike plus 31 from month 3's smaller cohort is 195 chargebacks, measured against month 3's 8,000 sales. That is 2.44%. More than 100 chargebacks with a ratio between 1.50% and 2.99% is the Excessive Chargeback Merchant band, and getting back out of it takes 3 consecutive months under both thresholds, not 1 good month.

Now run it again with the campaign still live. If month 3 had also done 18,000 sales, the same 195 chargebacks in month 4 come out at 1.08%, under the line. The disputes did not change. The denominator did. Pausing the campaign is what pulled the trigger.

That is not an argument for spending your way out. It buys a clean month and delivers a bigger cohort of disputes 60 days later, into a month with no budget behind it. The same lag distorts every dispute number you read while ramping, which is the subject of why your dispute rate is really reporting last month's ad spend.

• Month 1, normal trading: 5,000 Mastercard sales.
• Month 2, the campaign lands: 18,000 Mastercard sales.
• Month 3, campaign paused, still digging out of the fulfillment backlog: 8,000 Mastercard sales.
• Month 4, back to baseline: 5,200 Mastercard sales.

The 120 day clock starts at delivery, not at checkout

Visa's dispute window is 120 days, and it runs from the transaction date, the expected delivery date, or the service date, whichever applies. On a product that ships in 3 days that distinction is academic. On a campaign that outran your stock it is the whole problem.

A spike breaks fulfillment. Inventory runs out, the 3PL falls behind, ship times go from 3 days to 3 weeks, and a chunk of the cohort gets a delivery estimate pushed out a second time. Every push moves the start of the dispute window with it. Where no delivery date was ever given, the issuer waits 15 days after the sale before it can file. For genuine future delivery orders the outer boundary stretches as far as 540 days from the purchase date.

So the cohort you bought in March is still producing chargebacks in September, landing in months where your sales count went back to normal long ago. You cannot outrun that tail, and closing the account does not end it. An acquirer must file a termination listing when the criteria are met even if the qualifying month falls after the relationship is over.

Your acquirer saw the deterioration before you did

By the time a chargeback shows up in your dashboard the decision has usually been made, because chargebacks are not the first thing that leaves your account.

When a cardholder tells their issuer a charge was fraudulent, the issuer files a fraud report. On Visa that is a TC40. It reaches your acquirer whether or not that cardholder ever files a formal dispute, and it gets there well ahead of the chargeback. Risk is reading a fraud curve that your payments dashboard does not draw.

Visa's monitoring ratio counts both sides of that. The numerator is TC40 fraud reports plus TC15 disputes, over settled card not present transactions, all by count rather than by dollar value. The same fraud claim can appear as both, so it can move the ratio twice.

Settled is the word that matters. Declines are not in the denominator, so you can tighten authorization rules until you are turning away good orders and the ratio still will not move: the disputes you already sold sit in the top of the fraction while the bottom shrinks.

That is why the call comes before the chargebacks do, and why the first email from risk asks for tracking numbers, refund policy screenshots and a fulfillment timeline, not for your view on the ratio.

The volume itself is a breach of the agreement

There is an approved monthly processing volume on your merchant agreement and most operators have never read theirs. It is an underwriting decision, not an estimate. The acquirer is liable for refunds and chargebacks if you stop trading, so that number is the size of the exposure they agreed to carry.

A 3x month hands them 3x that exposure without asking. What follows is not punishment, it is a risk team pulling back an exposure nobody signed off:

If the freeze sits at a processor rather than a traditional acquirer the shape is the same and the clock is tighter, which is why what to actually do in the first 14 days of a payouts freeze is worth reading before you answer anything.

• Settlement stops. Daily batches move to manual review while somebody reads your fulfillment data.
• A reserve appears, or an existing one goes up. Rolling reserves in high risk commonly hold back 5% to 10% of each settled batch, and the hold runs 30 to 180 days before the oldest slice starts releasing.
• Payout timing stretches. Next day becomes weekly, weekly becomes held.
• Your approved volume gets rewritten downward, which caps the recovery you were counting on.
• If the account closes, funds already withheld can sit for the length of the reserve term, commonly up to 180 days.

The exact numbers, so you know where the lines sit

Learn these, because the person calling from risk will not read them to you.

A termination listing runs for 5 years and it carries the principal owner as well as the business: legal name, tax ID, address, phone number. The acquirer has to file it within 1 business day of terminating you, and only that acquirer can remove it, and only if it was filed in error. There is a single exception, the code for PCI non compliance, which comes off once you prove compliance. Excessive chargebacks is not that code, and winning the disputes afterwards does not undo the listing.

Look hard at reason code 04, because it has no minimum chargeback count. 6 chargebacks totaling $6,250 against 125 sales in a month clears both halves of it. A merchant with a high average order value can be listed on numbers that would never have triggered the monitoring program. Losing a processor costs you a few weeks. Being listed costs you underwriting everywhere for 5 years.

• Visa's acquirer monitoring program: on April 1, 2026 the excessive merchant threshold dropped from 220 basis points to 150, a 1.5% ratio, in the US, Canada, the EU, Asia Pacific and Latin America. CEMEA stays at 220. It only applies once you also record at least 1,500 combined fraud reports and disputes in a month, and merchants over the line are charged $8 per fraudulent or disputed transaction.
• Mastercard's Excessive Chargeback Merchant band: 100 to 299 chargebacks in a month and a ratio of 1.50% to 2.99%, measured against the previous month's sales count. Both conditions have to be met, and exit takes 3 consecutive months back under them.
• High Excessive: 300 or more chargebacks and 3.00% or higher. Fines escalate the longer you sit in either band, and an issuer recovery assessment can be charged on top.
• Mastercard's termination database, reason code 04: Mastercard chargebacks above 1% of monthly Mastercard sales that also total $5,000 or more, both in the same calendar month. Read that timing twice. Both halves sit in the same month here, so a collapsed sales month hits the listing test immediately rather than 30 days later.
• Visa's own screening database, reason code 22: 1,000 disputes and a 1.8% dispute to sales ratio in a single month, measured on amounts rather than counts.

What still moves the number inside 30 days

You have about a monitoring cycle to work with. Here is what each lever really does, including the part vendors leave out.

And the thing not to do: don't manufacture volume to hold the denominator up. It works for a month, then hands a larger dispute cohort into a month with no budget behind it. If you ramp again, ramp in steps your acquirer has been warned about, which is the unglamorous version of scaling Meta spend without breaking what sits downstream of it.

• Visa's rapid dispute resolution turns a qualifying dispute into an automatic refund before it becomes a chargeback. Here is the part that gets glossed over. Non fraud disputes resolved that way come out of the monitoring ratio, but fraud reports do not: a fraud claim settled through rapid dispute resolution or a network alert still counts against you, and the only route that scrubs the fraud report itself is winning a compelling evidence representment. If your cohort's problem is fraud claims rather than service complaints, this fixes less than the sales page suggests.
• Alert networks like Ethoca and Verifi buy you a refund window instead of a chargeback. You still eat the order value and the shipping, and coverage only reaches issuers inside the network, so a share of your disputes will never generate an alert.
• Refunding the worst slice of the cohort before they file. Unglamorous, and it works, because a refund never enters the numerator. Every unshipped order past its stated delivery estimate is a chargeback sitting on a timer.
• Fixing the billing descriptor and sending a delay email that names a new date. A large share of disputes off a cold viral cohort are people who do not recognize a brand name they saw once, in a feed, 5 weeks ago.
• Fighting the disputes you can win. Do that for the money, not for the ratio. Representment recovers funds. It does not remove the chargeback from the count that put you in the program, and it does not undo a listing.

The second failure nobody plans for

Something else breaks in this sequence, usually about 2 weeks after the payouts stop.

The card funding your Meta account is normally attached to the same business account the processor was settling into. When settlement stops, that balance drains against payroll and inventory, and the card Meta charges starts failing. Meta retries, the retry declines, and a failed billing attempt on an account already spending at spike levels is among the cleanest ways to get flagged. The billing signals that fire an unusual activity disable are almost all payment events rather than creative ones.

Then comes the move that ends it permanently. Somebody, often not the media buyer, looks at a Meta invoice for spend that produced revenue now sitting in a reserve, and disputes the charge. Do not. A chargeback against Meta becomes a debt owed to Meta, and it attaches to the assets connected to that payment method and that business. You can replace a merchant account. That one is far harder to walk back.

Which is why the funding path deserves as much thought as the account itself. The honest trade off with an agency account is real: you are inside somebody else's account, under their rules, and it can still be restricted like anything else on the platform. What changes is the plumbing. An account funded by topping up with crypto or a bank transfer is not sitting on the rail your processor just froze, so a payments problem stays a payments problem instead of becoming an advertising problem in the same 2 weeks. Replacements exist when an account does go down. And a flat monthly rate with no cut of spend does not scale with the spike, so the month that broke your processor does not also triple your infrastructure bill. If you are rebuilding, the spend ceilings on each tier matter first, because the tier has to fit the volume you intend to run next, not the volume that just got you here.