Definition
Chargeback
A cardholder dispute that forces the issuing bank to reverse a transaction through the card network, pulling funds from the merchant after the sale.
A chargeback is not a refund. It is a forced reversal of a card payment that the cardholder starts with their issuing bank, which then travels through the card network to the merchant's processor and finally hits your payout. The shopper may never open a ticket; they open a dispute. Money leaves your account, you pay a fee, and you usually have a short window to submit evidence or accept the loss.
In ecommerce, chargebacks cluster around late or missing parcels, unclear statement descriptors, product-not-as-described claims, and "I did not authorize this" when someone in the household did. Treat them as a payments and ops problem first, support problem second.
How a chargeback moves through banks and card networks
The flow starts with the cardholder, not with your helpdesk. The customer contacts the issuing bank and files a dispute under a reason code: fraud, merchandise not received, not as described, credit not processed, and similar categories that Visa, Mastercard, and other networks define. The issuer provisionally credits the cardholder and sends the dispute through the network to the acquirer that processes your store.
Your payment provider then pulls the disputed amount plus a chargeback fee from your balance or next settlement and opens a response deadline. That path is why a polite refund conversation and a chargeback feel different in the books. A refund is a merchant-initiated return of funds. A chargeback is a bank-forced pull with network rules, reason codes, and fees that do not vanish even if you later win.
Stripe's ecommerce chargebacks 101 walks through this lifecycle from the processor side. Operators who only monitor support tickets will miss the first signal until the payout dips or the processor email lands.
Friendly fraud, true fraud, and merchant error
Not every chargeback is a criminal card theft. True fraud is when a stolen card or credential-stuffed account places the order without the real cardholder's knowledge. Friendly fraud, also called first-party misuse, is when the legitimate cardholder disputes a purchase they authorized or received: "item never arrived" after a porch delivery, "not as described" after buyer's remorse, or a family member ordered and the cardholder denies it.
Merchant error is on you: wrong item shipped, double charge, failed refund, or a product page that overpromised. You need different responses for each bucket. True fraud wants tighter authorization, AVS/CVV checks, velocity rules, and sometimes 3-D Secure. Friendly fraud wants delivery proof, clear policies, and a support channel that is easier than calling the bank.
Merchant error wants root-cause fixes in fulfillment and billing so the same SKU or flow stops generating disputes. Blaming every case on "fraudsters" hides the operational leaks that keep your ratio high.
Fees, ratios, and what processors actually watch
Each chargeback usually costs more than the disputed order: a per-case fee, the product cost if you already shipped, and the time spent gathering evidence. Networks and processors also track chargeback ratios against sales volume. Cross program thresholds and you face higher reserve requirements, higher fees, forced 3-D Secure, or account termination.
That risk is why a spike of "item not received" disputes during peak season is an existential payments problem, not a minor CX annoyance. Watch the ratio by reason code, not only the total count. A sudden rise in fraud codes points to acquisition or bot traffic. A rise in merchandise-not-received points to carriers, tracking gaps, or long transit promises. A rise in not-as-described points to creative, sizing, or QC.
Finance should reconcile processor dispute reports with order IDs weekly so ops and support work the same list. Waiting for a monthly PDF is how stores discover they are already on a monitoring program.
Representment: the evidence packet that wins or loses
Representment is your formal response: you submit documents that match the reason code and argue the charge was valid. For merchandise not received, the core is carrier tracking with a delivery scan, shipping address matching the order, and dates that show the item left before the dispute.
For fraud claims on card-present-not-applicable ecommerce, you need AVS results, CVV checks, IP/device signals if available, customer communications, and proof of prior legitimate orders when relevant. For not-as-described, you need the product page as shown at purchase, photos of what shipped, and any return policy the customer ignored. Build a checklist per reason code and store it where anyone on the dispute queue can run it without inventing a process under deadline.
Include the order confirmation, invoice, tracking screenshot from the carrier (not only your app), delivery photo if the carrier provides one, refund or return history, and support transcripts. Incomplete packets lose even when you are right. If your 3PL or dropship supplier holds the only tracking trail, pull it into your system the day of shipment so representment does not depend on a slow email to a warehouse.
Prevention: descriptors, tracking, and support that is easy to reach
Most preventable chargebacks start with confusion, not malice. The statement descriptor on the card must match a brand name the customer recognizes; parent-company legal names and random DBAs drive "I do not recognize this charge" disputes. Ship with trackable methods for any order value that hurts to lose, email tracking links promptly, and surface live status in the account portal or chat so the shopper does not need the bank for reassurance.
Clear return and refund windows reduce "not as described" and "credit not processed" cases when customers can still solve the problem with you. Support quality matters, and it is not only AI. A bot that can look up the order and share tracking reduces WISMO friction; a dead-end bot that cannot reach a human after a failed delivery pushes people to the issuer.
Staff response SLAs on delivery exceptions, lost packages, and billing questions should be tighter than your refund window. Train agents to offer refunds or reships when policy allows before the dispute deadline. Prevention is descriptors plus logistics truth plus an accessible path to a person, not a single chatbot metric.
An operator playbook when a dispute hits
When a chargeback arrives, freeze the emotional response and run a fixed sequence. Tag the order, pull the reason code, and decide win probability within the first review: strong delivery proof and matching AVS is worth fighting; thin evidence on a low-value order may not be. If the product is still in transit, do not double-refund in panic and also fight the chargeback without coordination.
Document every decision in the order notes so the next person sees what was submitted. Feed wins and losses back into product and ops. Repeated "not received" on one carrier lane means change service level or address validation. Repeated descriptor disputes mean fix the billing text with your processor. Repeated friendly fraud on high-risk SKUs may justify signature required, holds for new customers, or stricter refund rules after delivery.
Chargebacks are expensive feedback. The stores that stay offline of network programs treat each reason code as a process defect until the data says otherwise.
Common questions
Frequently asked questions
What is a chargeback in ecommerce?
A chargeback is a cardholder dispute that reverses a transaction through the issuing bank and card network. The merchant loses the sale amount, usually pays a fee, and may submit evidence to contest the case.
How is a chargeback different from a refund?
A refund is merchant-initiated and stays between you and the customer. A chargeback is bank-forced, carries network reason codes and fees, and can affect your processor risk profile even if you later win.
What is friendly fraud?
Friendly fraud is when the legitimate cardholder disputes a purchase they authorized or received, often claiming non-receipt or non-recognition. It is different from stolen-card fraud and needs delivery proof and clear support paths more than fraud filters alone.
What evidence helps win a chargeback?
Match the reason code: tracking with delivery confirmation, AVS/CVV and order details for fraud claims, product page and shipping proof for quality claims, plus chat and email logs. Incomplete packets usually lose.
What prevents chargebacks besides AI support?
Use a clear card statement descriptor, send trackable shipments with proactive updates, publish return rules customers can use, and keep a fast human path for delivery and billing issues so calling the bank is not the easiest option.
Related terms
