Chargebacks sit in a blind spot because of where they land in the business. Refunds show up in Shopify. Ad spend shows up in the platform. Disputes show up in a payments dashboard that nobody opens, arrive 30 to 90 days after the order, and get written off as a cost of doing business. By the time anyone runs the number, the money has already gone and the ratio that governs your ability to take card payments at all has been quietly climbing.
Two things changed this year. Visa tightened the rules in April, and the composition of disputes shifted decisively towards the customer who actually placed the order. Both of those changes hit DTC harder than they hit anyone else, because card-not-present is all we do.
What a chargeback actually costs you
The order value is the smallest part of the bill. The 2026 LexisNexis True Cost of Fraud study puts the all-in figure at 5.13 for every 1 lost to chargebacks. That multiplier is not a scare statistic. It is arithmetic you can run on your own P&L.
Take a 60 order on a wellness brand with 30% contribution margin. The revenue reverses. The goods are gone, so you lose COGS again on the replacement sale you now need. The dispute fee lands at 15 to 25 depending on your processor. Outbound shipping and fulfilment are already spent. The acquisition cost that brought the customer in, call it 35, will never be repaid because that customer is not coming back. Then somebody on your team spends 40 minutes assembling evidence for a case that has a coin-flip chance of going your way.
At 30% contribution margin, that single dispute needs roughly ten clean orders behind it just to get you back to level. Run 40 disputes in a Q4 month and you have wiped out 400 orders of profit without a single line item on your dashboard saying so.
A refund costs you the margin on one order. A chargeback costs you the margin on ten, plus the acquisition spend, plus a tick on a ratio that decides whether you keep your merchant account.
The rules changed on 1 April 2026
Visa's Acquirer Monitoring Programme, VAMP, replaced the old split between fraud monitoring and dispute monitoring with one combined ratio. Fraud reports (TC40) and non-fraud disputes (TC15) are now measured together against your settled card-not-present transaction count. That single change matters more than the headline number, because a clean fraud record no longer offsets a messy dispute record. They go in the same bucket.
From 1 April 2026 the excessive merchant threshold dropped from 2.2% to 1.5% across the US, Canada, the EU, APAC and LATAM. CEMEA stays at 2.2%. Acquirers themselves have to hold under 0.7%, which is why your processor will act on you long before Visa does. For a US merchant, the excessive designation requires both a ratio above 1.5% and at least 1,500 fraud and dispute items in a calendar month. Breach it and the fees run at 8 USD per fraud and dispute item, with no warning tier to absorb the first month.
Card-not-present ecommerce typically runs between 0.6% and 1%. Under 0.9% keeps you clear of all four network thresholds with room to spare. If you are sitting at 1.2% and telling yourself you are fine because the old ceiling was 2.2%, you are now two-thirds of the way to a programme you do not want to be in, and the acquirer sitting between you and Visa has a far lower tolerance than the number you are measuring yourself against.
The dangerous pattern for DTC is seasonal. Q4 volume inflates the denominator in November and December, then the disputes from those orders land in January and February against a much smaller transaction base. Your ratio spikes in the quarter you can least afford a reserve.
Three quarters of your disputes are not fraud
First-party fraud, the polite name for a legitimate cardholder disputing a purchase they genuinely made, now drives roughly 75% of ecommerce disputes. A survey of 1,278 merchants across 37 countries found 64% reporting rising first-party misuse, with about a quarter seeing increases of 25% or more. Projections put friendly fraud at 28% of all chargebacks by 2031, up from 22% in 2026, with the transaction value climbing from 8.1 billion to nearly 16 billion USD.
This is why buying more fraud screening will not fix your problem. The card is real, the customer is real, the address matches, the device fingerprint is clean. Your fraud tool passes the order because the order deserves to pass. The dispute arrives six weeks later from a customer who forgot the purchase, did not recognise the line on their statement, could not find your refund policy, or discovered their bank's app makes disputing a charge easier than emailing you.
Which means the lever is not detection. The lever is clarity, communication, and making your own refund route the path of least resistance.
The four causes you can fix in a fortnight
I have audited this in enough brands to know the same four things show up nearly every time. None of them require new software. All of them are cheaper than the disputes they prevent.
Your billing descriptor does not match your brand
If the statement line shows your legal entity, a holding company, or an abbreviation nobody recognises, you are manufacturing disputes. A customer who cannot place the charge does not ring you, they ring their bank. Set the descriptor to the brand name on the packaging, add a contact number where your processor allows it, and make sure the order confirmation, the shipping notice, and the statement line all use the same words. This is a ten minute change in most payment settings and it is the single highest-return fix on the list.
Your delivery comms stop at the dispatch email
Most brands notify at label creation and then go silent. Item not received is one of the most common reason codes in DTC, and a meaningful share of it is parcels that did arrive, to a neighbour or a safe place, with nobody told. Send an out for delivery notice, a delivered confirmation, and a nudge three days later asking whether everything arrived. Each of those touches is a chance to resolve a problem in your inbox rather than in a dispute queue.
Subscription renewals arrive with no warning
Replenishment and subscribe and save programmes are the best revenue line in DTC and the worst dispute line. A charge the customer forgot was coming is a dispute waiting to happen. A pre-billing notice three days ahead, with the amount, the date, the contents, and a one-click skip, converts a percentage of those charges into deliberate purchases and the rest into skips instead of disputes. A skip costs you one cycle. A dispute costs you the customer, the fee, and a tick on the ratio.
Cancelling is harder than disputing
Put your cancellation and refund policy where a frustrated customer will find it in under fifteen seconds, not in the footer. Answer support messages inside a working day during peak. If the easiest route to their money back runs through your site rather than their banking app, most people will take it, because most people are not trying to steal from you. They are trying to solve a problem quickly.
Fighting disputes: what the win rates actually say
Merchants win an average of 43.8% of the disputes they represent. That sounds workable until you see the net recovery figure across all disputes, which sits near 10.7% once you account for second-cycle cases and the disputes nobody bothered to fight. Ad hoc manual responses, the kind where someone pastes an order confirmation into a form at the end of the month, win roughly 12% of the time.
The gap between 12% and the 70% to 85% that a properly run representment process achieves is not effort. It is structure. A generic order confirmation proves a sale happened. It does not rebut the specific claim made under the specific reason code. Item not received needs signed delivery or carrier tracking to the AVS-matched address. Not as described needs the product page as it appeared on the order date, the size or ingredient detail, and any support thread. Fraud-coded cases need device, IP, AVS and CVV match, plus prior undisputed order history from the same customer.
Be selective. Fraud-coded disputes are the hardest to overturn and often not worth the hours on a low-value order. Item not received and not as described are the defensible ones and they are also, conveniently, the ones your delivery comms and product page can be built to win in advance. Decide your threshold, fight everything above it with a reason-code template, and let the rest go rather than burning a person on cases you will lose.
Representment recovers money. Prevention protects the ratio. Winning a dispute still counts against your VAMP count in most cases, so a brand that wins every case and prevents none can still end up in a monitoring programme.
What this looks like in practice
A supplements brand I work with came to me about margin, not payments. Revenue was growing, contribution margin was drifting down by a point or two a month, and nobody could explain it. Their subscription base had roughly doubled over three quarters, which everyone treated as unqualified good news.
The dispute report told a different story. The bulk of their cases were subscription renewals, coded as unrecognised or unauthorised, from customers who had happily received three or four previous shipments. Their descriptor carried the parent company name. There was no pre-billing notice. Cancellation lived behind a support email with a two-day response time during peak.
We changed the descriptor, built a three-day pre-billing flow in Klaviyo with a one-click skip, moved cancellation into the customer account, and put a reason-code template behind their representment. Skips went up, which the founder hated for about a fortnight. Dispute volume came down, the fees came down with it, and the margin drift stopped. The skips were never lost revenue. They were disputes that had not happened yet.
Inside the system
How we build this for brands
The reason disputes go unnoticed is that nobody owns the number. We fix that with a reporting agent that reads live Shopify and payment data weekly and surfaces leakage and risk as a plain summary: dispute rate against the VAMP threshold, cost per dispute in contribution margin terms, and which reason codes are moving. The same agent flags the seasonal trap, the January ratio spike against a shrinking denominator, before it becomes a processor conversation. Alongside that, our VOC engine mines support messages and reviews for the exact language customers use when they are confused about a charge, which is where the descriptor and pre-billing copy comes from rather than from guesswork.
The prevention layer is lifecycle work. We build and deploy the flows in Klaviyo: pre-billing notices, delivery confirmation sequences, post-purchase check-ins that catch a problem while it is still an email, and win-back logic that treats a cancelled subscription as a retained customer rather than a lost one. Part of this runs live for portfolio brands today; the full system is what we deploy when we take a brand on.
Margin Audit
Find Out What Disputes Are Really Costing You
I will pull your dispute rate against the current network thresholds, break the cost down in contribution margin terms, and show you which reason codes are preventable before Q4 volume lands. Numbers and a plan, nothing else.
Book Your AuditFrequently asked questions
What is a good chargeback rate for an ecommerce brand?
Card-not-present ecommerce typically runs between 0.6% and 1%. Holding under 0.9% keeps you clear of all four card network thresholds with room to spare. Visa's VAMP excessive merchant threshold dropped from 2.2% to 1.5% on 1 April 2026 in the US, Canada, EU, APAC and LATAM, with CEMEA remaining at 2.2%, so the old comfort zone is gone.
What does a chargeback actually cost a DTC brand?
Far more than the order value. The 2026 LexisNexis True Cost of Fraud study puts the all-in cost at 5.13 for every 1 lost to chargebacks, once you include reversed revenue, lost goods, the dispute fee, shipping already spent, the acquisition cost that will never be repaid, and staff time. On a 60 order at 30% contribution margin you need roughly ten clean orders to cover one dispute.
What is first-party fraud and why does it matter more than stolen cards?
First-party fraud, often called friendly fraud, is when the legitimate cardholder disputes a purchase they actually made. It now accounts for roughly three quarters of ecommerce disputes. A survey of 1,278 merchants across 37 countries found 64% reporting increasing first-party misuse, with about a quarter seeing increases of 25% or more. Fraud screening cannot block it, because the transaction is genuine.
Is it worth fighting chargebacks or should I just accept them?
It depends on whether you fight them properly. Merchants win an average of 43.8% of the disputes they represent, but net recovery across all disputes sits near 10.7%. Ad hoc manual responses win roughly 12% of the time, while reason-code-specific evidence packets tied to a specific order ID perform far better. Item not received and not as described are the most defensible; fraud-coded disputes are the hardest to overturn.
How do I reduce chargebacks on Shopify before Q4?
Start with the four preventable causes. Make your billing descriptor match the brand name customers recognise. Send delivery communications past label creation. Give subscription renewals a pre-billing notice with a one-click skip. Make cancellation and refund easier to find than a bank dispute. Most brands cut dispute volume meaningfully in a fortnight without touching their fraud tooling.
What is Visa VAMP and how is the ratio calculated?
VAMP is the Visa Acquirer Monitoring Programme. It combines fraud reports (TC40) and non-fraud disputes (TC15) into a single ratio against settled card-not-present transactions, so a clean fraud record no longer offsets a high dispute count. From 1 April 2026 the excessive merchant threshold is a 1.5% ratio combined with at least 1,500 fraud and dispute items in a calendar month for US merchants, with fees of 8 USD per item and no warning tier.
About the author
Caner Veli is a DTC operator who has helped 350+ brands fix broken growth engines. He built Liquiproof from zero to 3,000+ global retailers in under 6 years. He now runs the same playbook, supported by AI systems he built himself, for DTC and CPG brands.