Now open · chargeback shield is liveSee what's new
All posts
Blog

The chargeback starts as a sentence

Visa cut the Excessive line to 1.5% on 1 April 2026, and monitoring programmes ignore whether you won. Only disputes resolved before they become chargebacks stay out of the maths.

Updated 1 August 2026

The threat arrives as a sentence, not as a transaction

There is a line like this sitting in a Shopify inbox right now.

"I have been waiting three weeks. I am done. I am calling my bank."

Nothing has happened yet. No money has moved, no fee has been taken, there is no case number and no evidence deadline. There is a customer, an order, and a sentence.

That sentence is the last moment you still control the outcome. Once the call is made, the decision leaves your store and moves to a bank, a card network and a set of forms. Before the call, it is still a customer you can satisfy.

Most support software does not treat that sentence as special. It becomes ticket 4813, queued behind a question about sizing.

The ground moved on 1 April 2026

On 1 April 2026, according to Visa's Acquirer Monitoring Program fact sheet, Visa cut the threshold at which a merchant is flagged Excessive from 2.2% to 1.5%, in the US, Canada, the EU and Asia-Pacific.

Put that in orders. At 1,000 orders a month the old line was 22 disputes. The new line is 15. Seven disputes that sat comfortably inside the limit in March are over it in April, with nothing about your store having changed.

Small stores feel it harder. At 400 orders a month the line is six disputes. Six. One bad shipping week from a supplier produces that. Stripe adds that the industry treats dispute activity above 0.75% as excessive, and that a sudden spike can flag you before you reach the formal threshold.

Winning does not protect you

Here is the fact that reorders everything else.

Stripe puts it plainly in its documentation: monitoring programs do not consider dispute outcomes. Every chargeback counts against your ratio. Including the ones you win.

Sit with that, because a category of software is sold on the opposite assumption. Representment tools gather evidence, file the response and win a share of cases back. That is real money recovered and worth having, but it does not move the number that gets you flagged. You can fight every dispute, win the ones you fight, and still cross 1.5%.

The waiting is not free either. Shopify charges a $15 chargeback fee in the United States, taken out of your payout immediately, before anyone has judged the case, refunded only if you win. You get 7 to 21 days to respond. So the money leaves, the fee leaves, you spend a week on evidence, you win, the money comes back, and the chargeback still counts.

What Visa does exclude

Now the other half of the same fact sheet, and this is the hinge of the argument.

Visa excludes disputes resolved through pre-dispute solutions from the ratio.

Resolve it before it becomes a chargeback and it never enters the maths. Not won. Not overturned. Absent.

Visa backs this enough to have built the tooling for it. It reports that Order Insight deflected 40 to 45% of confirmed first-party-misuse disputes by supplying purchase detail while the cardholder was still on the phone to their bank, from Verifi's reporting for October 2023 to September 2024. That is a save in the last thirty seconds, at the bank's end of the line.

Your email arrived days earlier.

The same holds one step down. An inquiry is not yet a chargeback, and no money moves during one. If nobody resolves it, it escalates, and only then do the funds and the fee leave. Shopify and Stripe both document that unescalated inquiries do not count toward monitoring programmes.

Where the existing tools sit on the timeline

The market is crowded, and these tools are good at what they do. The question is where each one sits.

ToolWhere it actsState of the dispute
Signifyd, Riskified, KountCheckout and fraud screeningDoes not exist yet, and the customer is not unhappy yet either
Chargeflow, DisputifierAfter the bank is involvedAlready filed, already counted
AidersThe email, before the bank is calledStill a customer, still yours to resolve

Chargeflow does read helpdesk tickets, which is a genuinely smart thing to do. It reads them after a dispute exists, to assemble the evidence that wins it. Everything in the first two rows acts on the transaction: at checkout, at screening, or once the bank has opened a case. By the time any of them can act, the chargeback exists, and it is counted whatever happens next.

Aiders sits earlier because it looks at a different object. The sentence in the email, not the transaction record.

More than a third of disputes are you being slow

The Sift Q4 2025 Digital Trust Index, run by Researchscape with 1,075 US adults in October 2025, breaks disputes down by cause. Missing or late deliveries are 17%. Delayed refunds are another 18%. More than a third of disputes are caused by a merchant being slow rather than a customer being dishonest.

Dishonesty is real. In the same survey 16% admitted filing a dispute they knew was false while happy with the purchase, and Visa estimates friendly fraud at around 20% of fraudulent disputes globally, up to 30% for high-volume online merchants. It is not the majority, and it is not the part you can fix this week.

The Chargebacks911 2025 Cardholder Dispute Index, covering more than 1,200 cardholders in the US and UK in July 2025, found that nearly half of consumers who dispute bypass the merchant entirely and go straight to the bank. You never hear from them. And 88% say winning one dispute makes them more likely to file another.

So half are out of reach. The other half email you first, and they are announcing it in writing.

What Aiders does with that sentence

Aiders reads the inbound email, checks the order in Shopify, and replies. Chargeback Shield is the part that watches for this.

It detects the threat while the customer is still emailing you. It catches the coded versions, because most people never use the word chargeback. They mention the bank. The lawyer. The resolution centre. The consumer authority. The vague "you leave me no choice".

It is built to be quiet, and that matters as much as the detection. Frustration is not a threat. Demanding a refund is not a threat. Asking for a manager is not a threat. A detector that fires on every annoyed customer gets muted within a week, and then the one real threat lands in a stream nobody reads. A noisy detector is worse than none. Only intent to go outside your store raises an alert.

Every alert carries the customer's exact words, the reasoning, a recommended action and a de-escalation strategy, on the order it relates to. Repeat threats from the same customer escalate into a case. From one alert you can refund in full, refund part, cancel the order, reply only, or mark it as not a real threat.

The safety properties are the point:

  • It only offers what it can deliver. A shipped order greys out cancel and gives the reason. An already refunded order greys out refund and gives the reason.
  • The refundable amount is what is genuinely left, reconciled against what has already been refunded, not the order total.
  • Nothing sends on its own. Drafting and sending are separate clicks, and money actions confirm on top.
  • The refund happens in Shopify first. The email only goes if the refund succeeded, so no customer reads "we have refunded you" for a refund that did not happen.
  • A double click cannot refund twice, and a customer with a recent chargeback cannot be handed more money by the agent.

The dropshipping case

If your fulfilment runs long, delivery delay is your dispute engine.

Stripe's guidance is to communicate shipping times before checkout, communicate delays quickly, offer a refund to customers who do not want to wait, and provide tracking. Stores rarely disagree with that. The failure is that the customer who does not want to wait writes on Saturday and the reply arrives on Tuesday.

Aiders answers from live order and carrier data, around the clock, in the language the customer wrote in, and can issue the refund itself into the original payment method. What stops it giving the store away is that your rules are enforced in code, not suggested:

  • A refund ladder. Each rung is an offer in the order you choose: a cash percentage, a full refund on return, store credit only. It cannot jump to a full refund without walking your rungs, and any rung can be offer but never pay.
  • A cash ceiling. Under it Aiders pays immediately. Over it the refund queues for your team, and the customer hears the same wording either way.
  • Per request type, one of three: Aiders handles it, Aiders drafts and you approve, or Aiders stops and alerts you.
  • By default it will not pay a customer to keep something they never received. A shipped but undelivered cancellation becomes a return on arrival.

You can rehearse any of it in a sandbox that runs the same logic with every real action simulated.

What doing nothing costs

Stripe documents Mastercard's ladder. An Excessive Chargeback Merchant is one with 100 or more chargebacks in a month and a rate of 1.5% or above. Month one costs nothing. Month two costs $1,000. It escalates to $100,000 a month by month 19, and $200,000 for the high-excessive tier. From month four you also pay $5 for every chargeback above 300.

The exit condition is the part to read twice. You must stay under the threshold for three consecutive months to get out.

That is not a fine you pay and clear. It is a rolling test you have to pass three times in a row, on a ratio that counts every dispute filed against you, including the ones you win. Which leaves one lever. Fewer disputes filed. Not more disputes won.

Start where the threat starts

The sentence is already in your inbox. It has a timestamp, an order attached, and a customer who has not called anybody yet.

Install Aiders on Shopify. The first 100 emails are free, once, not per month. Spam does not count: only emails that get a reply land on the meter.