Chargeback evidence scenarios by dispute type
Planning ranges by reason code and authentication setup. Modeled from Visa network rules and industry benchmarks, not measured CertNode Reflex outcomes. Updated May 2026.
Why dispute outcomes vary so much
Two merchants with identical-looking chargebacks can land on opposite outcomes because the variance comes from authentication data captured at payment time, not the quality of the dispute response. A merchant who runs 3D Secure changes the liability posture for qualifying fraud disputes; a merchant who does not remains exposed even with the same evidence package, narrative, and submission tooling.
This page breaks down per-reason-code planning ranges across five evidence configurations: an unrepresented merchant (industry baseline), an automated submission with standard evidence, plus three incremental upgrades (3DS enabled, device fingerprint captured via collect.js, qualifying CE 3.0 priors). The fifth column shows the modeled ceiling when all of these are stacked.
Planning range by reason code and setup
| Stripe Reason Code | Industry baseline no defense | Automated submission CVC + Radar + network auth | + 3D Secure issuer liability shift | + Device fingerprint collect.js installed | + CE 3.0 priors ≥2 prior same-card charges | Full-stack ceiling everything stacked |
|---|---|---|---|---|---|---|
fraudulent (Visa 10.4) | ~12% | 22% | 75% | 35% | 50-60% | 85-92% |
product_not_received (Visa 13.1) | ~25% | 45% | 48% | 50% | 50% | 75-85% |
duplicate (Visa 12.6) | ~40% | 65% | 65% | 65% | 65% | 80-85% |
subscription_canceled (Visa 13.2) | ~20% | 40% | 42% | 55% | 50% | 70-75% |
product_unacceptable (Visa 13.3) | ~15% | 35% | 35% | 40% | 38% | 55-60% |
credit_not_processed (Visa 13.6) | ~30% | 55% | 55% | 55% | 55% | 75-80% |
unrecognized (Visa 10.5) | ~15% | 30% | 50% | 55% | 70% | 75% |
Numbers shown for Visa. Mastercard typically tracks 5% lower across all reason codes. Amex and Discover use different representment rules; rough parity assumed.
How to read this table
Industry baseline
What an unrepresented merchant wins on average. Sourced from Chargebacks911's 2024 Chargeback Field Report and Visa Risk Operations benchmarks. Most merchants without dispute tooling sit at 12-20% across the board because they miss deadlines, submit incomplete evidence, or write generic narratives that fail the reviewer's reason-code check.
Automated submission
What CertNode Reflex delivers out of the box on a merchant with no extra setup. The lift comes from automated CVC/AVS/risk-score capture, narrative generation tailored to the reason code, and 21+ Stripe evidence fields populated within 60 seconds of dispute creation. This is the floor — what every merchant gets simply by installing.
+ 3D Secure
The single biggest evidence lever. When the cardholder authenticates via 3D Secure, Visa Core Rules Section 11.4 shifts liability to the issuer for fraud disputes (Visa 10.4). That liability shift materially improves the merchant's evidence posture on qualifying charges. Smaller lift on non-fraud reason codes. Enable in your Stripe Radar Rules.
+ Device fingerprint
Captured via the CertNode collect.js script at the moment of payment. Records device fingerprint, IP address, and browser environment as pre-dispute evidence. Strengthens the "cardholder was present" argument on fraud disputes and the "customer kept using the service" argument on subscription-canceled disputes.
+ CE 3.0 priors
Visa Compelling Evidence 3.0 qualifies a fraud dispute when the same cardholder has 2+ prior undisputed transactions on the same card with the same merchant, with matching IP or device. Reflex auto-detects qualifying scenarios and includes the CE 3.0 evidence block in submissions. Mostly benefits merchants with repeat customers.
Full-stack ceiling
All previous layers stacked. The highest fraud-dispute scenario is still a modeled ceiling; no chargeback is guaranteed to win. Card networks reserve discretion in close cases, and below ~$50 disputes get reflexively resolved in the cardholder's favor regardless of evidence weight.
Modifiers not in the table
Dispute amount under $50
Card networks reflexively favor the cardholder on small disputes regardless of evidence quality. Subtract roughly 10% across all rows when the disputed amount is below $50. Above $200, evidence weight starts mattering more and reviewers spend longer per case.
Time since purchase
Disputes filed more than 60 days after the original purchase carry less cardholder credibility — banks recognize the pattern of cardholder remorse versus genuine fraud. Slight uplift for the merchant on long-delay disputes if the merchant has activity records covering the gap.
Wallet authentication
Apple Pay, Google Pay, and Samsung Pay payments include biometric authentication on the cardholder's personal device. This is treated as stronger than 3D Secure by reviewers. When a fraud dispute hits a wallet-authenticated payment, the merchant typically wins by default.
Card brand
Visa figures shown above. Mastercard tracks roughly 5% lower across all reason codes. Amex disputes have a different process (issuer-mediated retrievals before formal chargeback) — Reflex's submission floor applies but the upside ceilings differ.
Which reason codes you'll actually see
Reason-code distribution varies by merchant category, but for most digital-services and small-ticket e-commerce merchants, the rough distribution is:
- ~70% fraudulent — by far the dominant category. The 3D Secure column moves more revenue than every other lever combined for most merchants.
- ~20% product_not_received — common for shipped goods. Delivery proof (Shippo, USPS, UPS API) is the lever.
- ~5% subscription_canceled / product_unacceptable — common for SaaS and subscription products. Cancellation policy + post-cancel login activity matter most.
- ~5% duplicate / credit_not_processed / unrecognized — handled cleanly by transaction records and refund history.
If your category skews differently (high-ticket, B2B, regulated), pull your actual reason-code distribution before deciding which authentication levers to prioritize.
Sources and caveats
Sources cited
- Chargebacks911 2024 Chargeback Field Report (industry baseline rates)
- Visa Core Rules and Visa Product and Service Rules (liability shift, CE 3.0 eligibility)
- Mastercard Chargeback Guide (representment requirements)
- Federal Rules of Evidence 902(13) and 902(14) (electronic records authentication)
- RFC 3161 (Internet X.509 Public Key Infrastructure Time-Stamp Protocol)
- CertNode internal scoring system (lift estimates per evidence factor)
What these numbers are NOT
Modeled outcome ranges from industry benchmarks plus internal scoring weights. They are not measured Reflex outcomes — Reflex is early-stage with too few resolved disputes to publish a meaningful measured rate. When we have 100+ resolved disputes per reason code, we will publish those measured rates here alongside the modeled ones.
Why not single-number outcome claims
Most chargeback-defense vendors publish single-number win-rate claims like "75% win rate" without per-reason-code or per-setup breakdowns. That number is misleading because it hides the underlying mix — a merchant who runs 3D Secure and a merchant who doesn't will see wildly different outcomes with the same vendor. The breakdown above lets you reason about your expected outcome from your actual setup, which is more useful than a brand-level average.
When this page was last updated
May 2026. Industry baselines refreshed annually. Visa CE 3.0 thresholds refreshed when Visa publishes rule updates. Internal scoring weights refreshed continuously based on observed outcomes.
Where this routes you
Three concrete moves based on what you saw above:
Install Reflex
Get the "automated submission" column lift on every dispute. 60-second evidence submission, AI narrative, 21+ Stripe fields populated.
Install Evidence Vault
Capture device fingerprint and IP at payment time. Pre-dispute evidence Reflex feeds into every dispute response.
Enable 3D Secure
In your Stripe Dashboard, add Radar Rules to request 3DS on elevated-risk charges. The biggest single lever in this table.