Chargebacks sit at the awkward intersection of customer service, fraud, payments engineering, and accounting. When a cardholder disputes a transaction, the issuing bank does not call the merchant first. It pulls the funds, files a reason code, and starts a clock. The merchant finds out later, often through a payment processor dashboard, and has a narrow window to respond. For US retailers and direct-to-consumer brands, the cost of getting this wrong is measured in real revenue, not just operational pain.
This guide walks through how card networks actually run the chargeback process in 2026, what each reason code typically means, where merchants lose disputes they could have won, and the practical playbook a finance or payments lead should put in place. It pairs with our pillar on how retail payments are changing across cards, BNPL and crypto, which frames where chargebacks fit inside the wider payments stack.
In short
- Chargebacks are network-driven reversals, not regular refunds. Visa, Mastercard, American Express, and Discover each run their own dispute rules, but they share the same broad arc: dispute, representment, pre-arbitration, arbitration.
- Reason codes matter more than the dollar amount. Fraud, “item not received,” and “not as described” each demand different evidence, and merchants who treat them the same lose winnable cases.
- Most chargebacks are won or lost before they happen. Clear billing descriptors, real-time fraud scoring, delivery tracking, and a documented return policy decide outcomes more than slick representment letters.
- Visa Compelling Evidence 3.0 and Mastercard Collaboration changed the math. If you can prove a prior undisputed transaction from the same cardholder, many fraud disputes never reach representment in the first place.
- Track the ratio, not just the count. Networks act on chargeback-to-transaction ratios. Crossing thresholds triggers monitoring programs that carry per-dispute fines and, eventually, the loss of card acceptance.
Why card network chargebacks merchant exposure is rising in 2026
Three forces have pushed dispute volume up over the past two years. The first is the long tail of “friendly fraud,” where a real cardholder buys something, receives it, then disputes the charge anyway. Industry trackers consistently put friendly fraud at well over half of all e-commerce disputes, and the rate climbed again through 2025 as households tightened budgets. The second force is the rise of buy now pay later and subscription billing, both of which generate complaints about installments, free trials, and cancellation flows that consumers route through their card issuer instead of the merchant. The third is the slow but real consolidation of network dispute rules, which has made it easier for issuers to file disputes by code and harder for merchants to win by improvisation.
For a small store running on a single payment processor, the visible symptom is the same: more disputes per thousand orders, more processor fees layered on top of the lost transaction, and more pressure from the acquirer to bring the ratio down. The economics get worse fast. A single $80 dispute often costs the merchant $80 in returned funds, a $15 to $25 chargeback fee, the cost of goods, and any shipping. Win the representment and most of that comes back. Lose, or fail to respond, and it stays gone.
What a chargeback actually is, in network terms
A chargeback is a forced reversal of a card transaction initiated by the cardholder through their issuing bank. It is not a refund. A refund is voluntary and routed through the merchant. A chargeback is mandatory once filed, and the funds move first while the dispute is resolved later. The four major US card networks (Visa, Mastercard, American Express, Discover) each set the rules, but their flows rhyme. Our deeper explainer in how card networks really work behind every retail checkout walks through the routing layer that sits beneath every dispute.
The dispute is filed under a reason code. Visa uses a four-category schema (fraud, authorization, processing errors, consumer disputes). Mastercard uses similar buckets under its Mastercom platform. American Express, which is both network and issuer for most of its volume, runs its own reason code list. Discover is the smallest of the four but matches Visa fairly closely. The reason code drives everything: what evidence is admissible, how long the merchant has to respond, and whether the network will accept a representment at all.
The four phases every merchant should recognize
- Retrieval or pre-dispute inquiry. Some networks, mainly Visa via Order Insight and Verifi, push a request for transaction detail before the formal dispute. Answer these. They often resolve the case before it becomes a chargeback.
- First chargeback. The issuer pulls the funds and files the reason code. The merchant has between 20 and 45 days to respond depending on network and code.
- Representment. The merchant submits evidence. If the issuer accepts it, funds return. If not, the case can escalate.
- Pre-arbitration and arbitration. The issuer can push back with new information. Arbitration filings carry significant fees (typically $500 or more if the merchant loses) and are reserved for cases worth fighting on principle or precedent.
How the four major networks compare on dispute rules
The differences look small in a policy document and large in a finance review. The table below summarizes the practical contours US merchants should track in 2026.
| Network | Merchant response window | Headline reason buckets | Notable 2024–2026 changes |
|---|---|---|---|
| Visa | 30 days (most codes) | Fraud, Authorization, Processing, Consumer | Compelling Evidence 3.0 lets merchants block fraud disputes using prior undisputed transactions |
| Mastercard | 45 days (most codes) | Authorization, Cardholder, Fraud, Point of Interaction | Collaboration platform expanded; first-party fraud reason code 4870 in active use |
| American Express | 20 days (issuer and network combined) | Fraud, Inquiry/Miscellaneous, Authorization, Processing, Card Member Dispute | Tighter SafeKey 2.0 requirements for digital fraud cases |
| Discover | 30 days | Fraud, Authorization, Processing, Cardholder Dispute | Aligned more closely with Visa post-2024 acquisition-era rule harmonization |
The single most important number on this table is the response window. American Express in particular runs the tightest clock, and merchants who outsource dispute handling to a third party without monitoring the queue lose Amex cases more often than any other network simply by missing the deadline. For a broader survey of vendors that automate this layer, see our review of tools and vendors for card networks in 2026.
The reason codes that drive most US disputes
About 80 percent of merchant chargeback volume sits in a handful of codes. Knowing them by heart, and what each one requires, is the difference between winning 60 percent of representments and winning 20 percent.
Fraud codes (Visa 10.4, Mastercard 4837, Amex F29/F30)
The cardholder, or someone claiming to be the cardholder, says the transaction was not authorized. This is the category where Visa’s Compelling Evidence 3.0 reshaped the playing field. If the merchant can show two prior undisputed transactions from the same cardholder, on the same card, with matching device, IP, or shipping data, Visa will block the dispute from progressing. Mastercard reason code 4870 covers a similar first-party fraud scenario but with stricter evidence requirements.
The practical lesson is that fraud disputes are often won at the data layer, not the letter-writing layer. A merchant that retains device fingerprints, hashed IP records, and matching customer history across orders is in a different position from one that only stores order totals.
“Item not received” (Visa 13.1, Mastercard 4855)
The cardholder says the goods or services never arrived. The merchant defense is delivery evidence, ideally signed proof of delivery for orders above the carrier signature threshold (typically $750 for major US carriers, though FedEx and UPS allow merchants to set lower thresholds). Tracking numbers alone are weak evidence. Photo-on-delivery, signature capture, and matching delivery address to the billing record are strong.
“Not as described” (Visa 13.3, Mastercard 4853)
The cardholder received the item but says it does not match the listing. This is the hardest category to win because the network often defers to the cardholder’s description. The merchant defense is the product listing as it appeared at the time of purchase, customer service correspondence, and any return policy the cardholder agreed to at checkout. Archive product pages and terms of service with timestamps. Merchants who rebuild their site cannot prove what the listing actually said on the day the order was placed.
Subscription and recurring billing (Visa 13.2, Mastercard 4841)
The cardholder claims they canceled or did not authorize the recurring charge. The defense is the original subscription consent, a record of cancellation flow access, and clear billing descriptors. Networks have been increasingly aggressive in pushing merchants to honor cancellation requests promptly and to send pre-renewal notifications. The Federal Trade Commission’s “click to cancel” rule, with its 2024 finalization and 2025 enforcement ramp, sits underneath these disputes.
Common mistakes that cost merchants winnable cases
Most lost disputes are not lost on the merits. They are lost because the merchant did one of a small number of recoverable things wrong. The pattern repeats across industries.
- Treating the chargeback like a refund. Issuing a separate refund after a chargeback is filed gives the cardholder the funds twice and concedes the dispute. Reverse the chargeback through representment, do not parallel-refund.
- Generic representment letters. A boilerplate response that does not address the specific reason code is rejected almost automatically. Each code has a checklist; use it.
- Ignoring retrieval requests. Visa Order Insight and Verifi requests are not chargebacks yet. They are a chance to provide context that may stop the dispute. Merchants who treat them as low priority watch dispute volume rise.
- Sloppy billing descriptors. A descriptor that does not match the customer’s expectation (the brand name they bought from, not a holding company) triggers “I do not recognize this charge” disputes that did not need to happen.
- Not monitoring the dispute ratio. Crossing 0.9 percent of transactions or 100 disputes per month puts a merchant into Visa’s Dispute Monitoring Program. Crossing 1.8 percent puts them into the High-Risk version, which is hard to exit and expensive while it lasts.
- Late responses on Amex. The 20-day window is half of what most merchants expect, and Amex does not extend it for processor lag.
- No internal owner. Disputes that bounce between finance, customer service, and operations get missed. One named owner, with a queue and an SLA, wins more cases than three teams reacting ad hoc.
Examples from US retail and e-commerce
The pattern is easier to see through real-world shapes that show up across categories.
Apparel and beauty. A mid-size DTC apparel brand running on Shopify with Stripe sees its dispute ratio creep from 0.4 percent to 1.1 percent over two quarters. The driver is not fraud; it is “item not received” disputes on orders shipped through a regional carrier that does not capture signature. Switching to signature-on-delivery for orders above $150 and adding photo-on-delivery for the rest drops the ratio back below 0.6 percent within 60 days. The cost of the signature upgrade is a fraction of the chargeback fees it prevents.
Electronics resellers. A consumer electronics reseller listing refurbished phones sees a wave of “not as described” disputes after a packaging redesign. The new boxes do not clearly mark “refurbished,” and cardholders dispute on the grounds they expected new product. The win rate on these disputes hovers around 20 percent because the listing language is ambiguous. Reverting the box copy and adding a checkout confirmation step (“I understand this is a refurbished device”) cuts the dispute rate in half over the next two months.
Subscription box services. A monthly subscription box company with $25 average order value runs into the FTC click-to-cancel rule mid-2025. Their cancellation flow requires a phone call during business hours. Disputes spike. The company rebuilds the flow to allow web-based cancellation matching the signup path, and dispute volume drops by 40 percent within three billing cycles.
Marketplace operators. A US marketplace platform with 4,000 third-party sellers handles disputes centrally but does not chargeback the responsible seller for losses. When the dispute ratio approaches the Visa monitoring threshold, finance pushes through a contract amendment that allocates the chargeback liability to the seller of record. Sellers in turn tighten their listings and fraud controls. The platform’s net loss falls sharply, and the underlying seller behavior improves.
The 2026 representment playbook
For merchants who want a single workflow, the steps below collapse the common patterns into something a small payments team can actually run.
- Receive the dispute. Pull the reason code, dispute amount, transaction date, and the issuer’s stated reason from the processor dashboard.
- Classify within 24 hours. Decide one of three paths: accept (cost of fighting exceeds dispute value), represent (evidence supports merchant), or escalate to arbitration consideration (precedent matters).
- Pull evidence by code. For fraud, pull device, IP, shipping address match, and prior transaction history. For “item not received,” pull delivery proof. For “not as described,” pull the listing snapshot and customer service log.
- Write a code-specific cover letter. Three paragraphs maximum: what the transaction was, why the cardholder’s claim does not match the evidence, what the evidence proves. Avoid emotional language. Networks read these for facts, not narrative.
- Submit before the deadline minus three business days. Build a buffer for processor lag.
- Track outcome and feed back. Every win and loss should update the representment template for that code. After 90 days, your win rate by code should be visible on a single dashboard.
How chargebacks affect business valuation and exit conversations
Chargeback ratios show up in due diligence packs. Acquirers of US retail and e-commerce businesses ask for 24 months of dispute history, broken down by reason code, network, and processor. A dispute ratio above 1 percent flags risk because it implies either weak fraud controls, weak post-sale operations, or both. A clean dispute history adds defensible margin to the multiple. For sellers thinking about the longer arc, our piece on valuation methods that buyers actually use for retail businesses covers how operational metrics like this feed into the headline number.
For sellers, the practical implication is to clean up disputes well before a process starts. Six months of disciplined dispute management changes the picture meaningfully. The ratio drops, the reason code mix becomes legible, and the diligence narrative is straightforward.
Tools and partners worth knowing
The chargeback management category sits beneath the broader card-network tooling stack we covered in tools and vendors for card networks in 2026. The tools that matter for dispute work fall into three groups.
Pre-dispute deflection. Verifi (a Visa company) and Ethoca (a Mastercard company) sit between the issuer and the merchant. They route inquiries before they become disputes. Most processors integrate one or both by default; the question is whether the merchant is actually responding to the inquiries that come through.
Representment automation. Chargebacks911, Chargehound, Justt, and Kount Dispute Management automate evidence collection and submission across networks. For merchants above 50 disputes per month, the labor savings usually justify the per-dispute fee. Below that volume, manual handling is often cheaper.
Fraud and friendly fraud detection. Signifyd, Riskified, and Forter sell guarantee-backed fraud screening that includes a chargeback indemnification component for approved orders. The economics work best for high-AOV businesses where a single missed fraud chargeback is meaningful.
Working with the wider payments stack
Disputes do not happen in isolation. The same systems that authorize a transaction shape what evidence is available when the dispute lands months later. Tokenization, 3-D Secure flows, address verification (AVS), and card verification value (CVV) checks all leave artifacts that strengthen representment cases. Merchants who treat payments as a black box maintained by their processor lose visibility into these signals. Merchants who instrument their checkout (storing the authentication outcome, the AVS response, the device fingerprint, and the customer history) win disputes that look unwinnable on the surface. The pillar piece on how retail payments are changing across cards, BNPL and crypto spends more time on this layer, including how BNPL providers handle their own disputes outside the card network rails.
For deeper context on how the network rails themselves move money, the explainer at the Federal Reserve Payments page is a useful primer on the broader US payments system that sits underneath card disputes.
What to measure every month
Five numbers belong on the monthly payments dashboard. They are not the only metrics that matter, but they are the ones that move first when something is wrong and the ones acquirers and processors look at first.
- Dispute ratio by network. Track Visa, Mastercard, Amex, and Discover separately. Network ratios are calculated separately, and so should monitoring.
- Reason code mix. A shift toward “not as described” suggests product or listing issues. A shift toward fraud codes suggests checkout or fraud screening issues.
- Representment win rate. Overall and by code. Anything below 40 percent on a code suggests the evidence pack is weak.
- Average dispute amount. A rising average suggests larger orders are being targeted; tighter fraud screening on high-AOV orders may be warranted.
- Time-to-respond. Track the median and the 95th percentile. The 95th percentile is where missed deadlines hide.
FAQ
What is the difference between a chargeback and a refund?
A refund is a voluntary reversal initiated by the merchant. A chargeback is a forced reversal initiated by the cardholder through their issuing bank, with the funds pulled first and the dispute settled afterward. Refunds carry no chargeback fee and no impact on the dispute ratio. Chargebacks carry both.
How long does a chargeback take to resolve?
From dispute filing to final resolution, a typical case takes 60 to 90 days. Representment usually returns funds within 30 days of submission, but the issuer can escalate, which extends the timeline. Arbitration cases can take six months or more.
What is the chargeback threshold that triggers monitoring programs?
Visa places merchants into its Dispute Monitoring Program at a dispute ratio of 0.9 percent and 100 disputes per month, and into the High-Risk version at 1.8 percent and 1,000 disputes per month. Mastercard’s Excessive Chargeback Program triggers at 1.5 percent with at least 100 disputes per month. Amex and Discover have similar thresholds but enforce them through account-by-account reviews rather than published programs.
Can a merchant prevent friendly fraud?
Not entirely, but a clear billing descriptor, prompt customer service response, easy refund flow, and prior-transaction evidence under Visa Compelling Evidence 3.0 prevent a significant share of friendly fraud disputes from succeeding. The goal is to make the dispute either unnecessary (the customer reaches you first) or losable for the cardholder (the evidence is overwhelming).
What evidence wins “item not received” disputes?
Signed proof of delivery, photo-on-delivery, and a delivery address that matches the billing address. Tracking numbers alone are weak. For orders above $750 in the US, signature is the carrier default for major couriers and should be required by the merchant for high-AOV orders below that.
Should small merchants use representment automation tools?
Below roughly 50 disputes per month, manual handling is usually cheaper. Between 50 and 500 disputes per month, automation pays off in labor savings and consistency. Above 500, automation is essentially mandatory because no manual team can keep up with the deadlines across networks.
Does 3-D Secure authentication transfer chargeback liability?
For most fraud-related reason codes, yes. When a transaction is fully 3-D Secure authenticated (Visa Secure, Mastercard Identity Check, Amex SafeKey, Discover ProtectBuy), the liability for fraud chargebacks shifts to the issuer. Friendly fraud and non-fraud reason codes are not covered by the shift.
How do BNPL and digital wallet disputes work?
BNPL providers (Affirm, Klarna, Afterpay) handle disputes outside the card network rails. The merchant typically faces the BNPL provider rather than an issuing bank. Digital wallets like Apple Pay and Google Pay route disputes through the underlying card network, so the same Visa, Mastercard, Amex, or Discover rules apply.
Bottom line for US merchants in 2026
Chargebacks are not a back-office nuisance. They are a leading indicator of where the operation is leaking trust, evidence, or attention. The merchants who handle them well do three things consistently: they capture the data that wins disputes at the moment of authorization, they classify and respond by reason code within tight internal deadlines, and they treat the dispute ratio as a board-level metric rather than a processor problem. The networks have given merchants more tools in 2024 to 2026 than at any prior point. Using those tools well is now the differentiator. For the broader payments picture these disputes sit inside, the place to keep reading is the pillar guide to how retail payments are changing across cards, BNPL and crypto.