Guides
What Evidence Actually Wins a Chargeback Representment (And What Loses)
Representment win rates swing from 15% to 60% depending on the dispute reason code and what you put in the evidence packet. Here's the breakdown, by reason code, of what actually moves an arbitration decision, and the packet contents that get thrown out.
When a chargeback lands in your Shopify Payments inbox, you have a deadline (usually 7-20 days depending on the processor) to submit representment. Evidence goes in, a card network reviews it, and either you win the money back or you don't. The outcome of that arbitration is decided by what's in the packet.
The unhelpful truth most merchants discover after their first ten disputes: the evidence that wins one kind of dispute is completely different from the evidence that wins another. A delivery photo that will win an "item not received" dispute is barely relevant to a "not as described" dispute. A device fingerprint that nails a "not authorized" fraud dispute is irrelevant for "credit not processed." Stuffing everything you have into every packet isn't thoroughness. It's noise, and it actively hurts your win rate because reviewers stop looking once the relevance drops.
This post is the reason-code-specific breakdown. What actually wins, what actually loses, organized by the dispute types you'll see most often on Shopify Payments.
The reason codes that matter
Card networks publish dozens of chargeback reason codes. Most Shopify merchants will see their dispute volume concentrate in five categories:
- Fraud / not authorized (Visa 10.4, Mastercard 4837): "I didn't make this purchase."
- Item not received (Visa 13.1, Mastercard 4855): "The package never arrived."
- Product unacceptable / not as described (Visa 13.3, Mastercard 4853): "The item wasn't what was advertised."
- Credit not processed (Visa 13.6, Mastercard 4860): "I returned it and never got my refund."
- Duplicate processing (Visa 12.6, Mastercard 4834): "I was charged twice for one order."
Each has a different winnable-with-what structure. Let's walk through them.
Fraud / not authorized
What the customer is claiming. Someone used their card without permission. They want a full refund.
Win rate with strong packet. 50-70% when the packet is on point. 20-30% with generic packets.
What wins:
- Device fingerprint matching prior legitimate transactions from the same customer. If the device used for this "unauthorized" transaction matches devices used for prior undisputed transactions on the same account, the fraud claim collapses. This is the single strongest signal.
- IP address consistency. Same network, same region, same browser fingerprint as prior purchases = very hard to argue this was a stranger.
- Delivery to the cardholder's address on file. If the shipment went to the billing address, fraud claims are much harder to sustain. Goods shipped to a random address are the opposite: they make the fraud claim stronger.
- AVS and CVV match at authorization. Matching AVS and CVV at the original transaction is a baseline requirement; without it, most fraud disputes are effectively conceded.
- Post-purchase behavior that implicates the cardholder. Login from their device after the purchase. Account modifications (password change, address update) in the same session. Delivery confirmation to their phone via SMS they responded to.
What loses (or is irrelevant):
- Proof of delivery alone. If the claim is "I didn't authorize this," showing the package arrived at an address doesn't help. You need to show it arrived at their address.
- Screenshots of the product listing. Completely irrelevant to whether the transaction was authorized.
- The customer's previous purchase history alone. Prior purchases don't prove this specific one was authorized. But prior purchases + matching device fingerprint = together they win.
- "We have no record of fraud complaints from this customer." Not evidence.
The pattern: fraud disputes are won by identity evidence, not purchase evidence.
Item not received
What the customer is claiming. They paid, package didn't arrive. Refund please.
Win rate with strong packet. 60-75% when delivery proof is real. Under 20% without it.
What wins:
- Delivery confirmation with address detail. Not just "delivered", but the carrier event log showing delivery to the exact shipping address on the order.
- Photo proof of delivery. FedEx, UPS, and increasingly USPS provide driver-taken photos of packages on porches. This evidence is almost uncontestable when available.
- Signature required + signature capture. If you required a signature and have the signature record, INR disputes almost always lose. This is why signature confirmation on high-ticket orders pays for itself, not for the delivery confirmation itself, but for the chargeback defense.
- GPS coordinates on delivery scans. Some carriers (UPS, sometimes FedEx) include GPS on the delivery scan. Matching GPS to the shipping address is strong evidence.
- Prior successful deliveries to the same address. Evidence that the address is real, receives mail, and the customer has received prior orders without issue.
- Time between order and dispute. A dispute filed 70 days after delivery on an order that had support contact at day 2 (customer confirmed receipt via email) is a losing dispute for the customer. Include support correspondence.
What loses:
- Tracking numbers with only "label created" or "in transit", no delivery scan. You have no defense and should probably concede.
- Delivery scans to a general area with no address detail. Some USPS scans are vague. Weak evidence.
- Carrier claim that the customer's building has "high theft rate." Not representment evidence.
- Refund policy screenshots. Irrelevant: the customer isn't claiming to have returned anything.
The pattern: INR disputes are won by delivery telemetry quality. Invest in carriers and methods that produce high-quality delivery evidence.
Product unacceptable / not as described
What the customer is claiming. They got the product, but it's defective, damaged, or not what was described.
Win rate with strong packet. 25-40%. This is the hardest dispute to win.
Why so hard: the card networks generally side with the cardholder on quality disputes because the merchant has no way to prove a subjective claim wrong from distance. The evidence that matters is mostly about the transaction terms, not the product condition.
What wins (what little does):
- Clear product listing, terms of sale, size charts. Screenshots of the product page the customer saw at time of purchase. If the customer's claim is "the color was different," a listing screenshot showing the exact color named in the dispute is helpful.
- Return was offered and refused (or ignored). Evidence that your return policy was available and the customer chose to dispute instead of returning. This is a significant factor: card networks expect customers to attempt resolution through the merchant first.
- Item was used / beyond return window. Evidence the customer kept and used the product for weeks before disputing.
- Customer correspondence accepting the product. An email thread where the customer acknowledged receipt and expressed satisfaction before filing the dispute.
- Verified third-party reviews of the same SKU. Limited weight but can support a "product as described" argument.
What loses:
- Your opinion that the product is fine. Doesn't matter.
- QC checklists signed by your fulfillment team. Weakly probative, not decisive.
- "Other customers have been happy with this product." Irrelevant.
- Industry reviews of your brand generally. Not transaction-specific, not helpful.
The pattern: "not as described" disputes are lost more than they're won. The best defense is preventive: photo documentation of outbound shipments for high-ticket items, clear policies that make the customer try to return before disputing. Once the dispute has landed, you're fighting uphill.
Credit not processed
What the customer is claiming. I returned this. You didn't refund me.
Win rate with strong packet. 50-70% if you're actually right. Near zero if you messed up.
What wins:
- Refund transaction record. The ledger entry showing you processed the refund, the date, and the amount.
- Notification to the customer. Email receipt of the refund, date/timestamped.
- Refund to a different payment method than the dispute. This happens: customer requested a refund to a different card, you processed it to the new card, they disputed the original. Clear documentation wins.
- Evidence the customer acknowledged the refund. A reply to your refund email, a support ticket referencing receipt of refund.
- No return received evidence (if your case is "they disputed without returning the item"). Tracking showing no return shipment. Policy screenshots showing return was required. Customer communication showing refund was contingent on return.
What loses:
- "We have a policy of always refunding within 5 days." Irrelevant without transaction-specific evidence.
- Partial refund claims without breakdown. If you issued a partial refund (restocking fee, shipping retained), you need to document the breakdown clearly. Vague evidence loses.
The pattern: this dispute is decided almost entirely by your records. If your refund records are clean and timestamped, you win. If they're missing or ambiguous, you lose.
Duplicate processing
What the customer is claiming. I was charged twice for one thing.
Win rate with strong packet. 70-85% when the charges are actually distinct orders.
What wins:
- Two distinct order records. Different order IDs, different items, different fulfillment dates.
- Different shipping tracking numbers. Evidence that two packages went to the customer, not one.
- Customer acknowledgement of both orders. Support tickets, order confirmation emails the customer opened.
- Different cart snapshots. If your system preserves cart contents, showing that the two transactions had different cart composition is decisive.
What loses:
- "We have no duplicate charge protection in our platform." Technical details don't matter; outcomes do.
- Missing order records for one of the charges. If you can't produce evidence for both transactions, you have a real duplicate and should concede.
The meta-pattern: relevance beats volume
The single most common representment mistake merchants make is dumping everything into the packet. Every email, every screenshot, every support ticket, every product photo. Reviewers don't read 80 pages of tangential material. They skim the first few pages, look for the evidence specific to the reason code, and decide.
A 5-page packet with the five most relevant pieces of evidence beats a 40-page packet with everything. Train yourself and your team to think by reason code:
- Fraud dispute → device and identity evidence first.
- INR dispute → delivery evidence first.
- Not-as-described → transaction terms and return-offered first.
- Credit not processed → refund ledger first.
- Duplicate → distinct-transaction evidence first.
Put the strongest evidence on page one. Put the rebuttal language specific to the customer's claim on page two. Supporting context on pages three to five. Stop there.
What RefundSentry produces for representment
One of the practical advantages of running a fraud scoring engine on your order flow is that the scoring signals double as representment evidence. Device fingerprint, IP address, prior purchase correlation, address history, return velocity: all of these are captured per order and per customer.
When a chargeback lands, the scoring record for the disputed order includes the signals you need for a fraud dispute packet. For INR disputes, the delivery telemetry pulled during enrichment can be included. For "credit not processed," the refund record on the order can be included. You're not recreating the evidence under deadline. It's already captured.
For high-volume merchants, this is the difference between a one-hour representment job and a two-day fire drill.
The takeaway
Representment isn't about having a lot of evidence. It's about having the right evidence for the specific dispute reason. A merchant who learns the five main reason codes and builds reason-specific packet templates will win 50-70% of their disputes. A merchant who throws everything in the folder wins 20-30%.
The evidence you need exists in your systems already. It just needs to be organized by reason code, surfaced at representment time, and formatted for the reviewer who has 90 seconds to decide. Structured once, it's a five-minute task per dispute. Unstructured, it's a half-day per dispute and a losing win rate.