Dispute and Return History
An account's and its linked entities' record of chargebacks, ACH returns, refunds, product returns and claims: a lagging but high-quality label that drives first-party fraud detection.
What it measures
| Reversal type | Key features |
|---|---|
| Card disputes | Count and rate by reason category (fraud vs. non-fraud); win/loss outcome; time from transaction to dispute; disputes after usage evidence (logins, downloads, deliveries) |
| ACH returns | Return codes by type (unauthorized/revoked authorization vs. insufficient funds vs. account closed or invalid); returns after value was withdrawn |
| Refunds and returns | Return rate by value and category; returns without receipt; returns of high-resale items; refund to a different instrument; partial returns that preserve incentives |
| Marketplace claims | INR/SNAD claim rate; claims vs. delivery evidence; claims across many sellers; claims escalated to the card network after platform resolution |
| Rewards reversals | Cashback or offer earnings later reversed because the purchase was returned or refunded |
| Linked history | Reversal history on the same device, instrument, bank account, address or identity across other accounts |
Why it matters
First-party fraud (the real customer reversing a legitimate transaction) is invisible to identity and device signals because the customer is who they say they are. History is the main evidence. A customer who has filed several "unauthorized" card disputes against merchants where their device logged in and their address received the goods has a pattern that one dispute never shows.
History is also the label source that trains everything else (Fraud Labeling and Feedback Loops). ACH unauthorized returns, fraud chargebacks and confirmed claim abuse are how you find out an earlier funding, purchase or payout decision was wrong. History also matters for revenue recovery now. Visa's Compelling Evidence 3.0, in force since April 2023, lets merchants defend card-not-present fraud disputes (reason code 10.4) with two prior undisputed transactions on the same card, 120 to 365 days old, sharing specific data elements. That makes clean history as valuable as bad history.
The pattern differs by industry. In Retail and Ecommerce it's serial returners, wardrobing and refund-to-different-card. In Online Marketplaces it's buyers with high INR/SNAD claim rates across sellers. In Cashback, Coupons and Rewards it's buy–claim–return against card-linked offers and receipt cashback. In Neobanks and Consumer Fintech and Brokerage and Investing it's ACH funding that comes back unauthorized after value has been withdrawn, and in Payments and P2P it's repeated P2P disputes.
How to collect it
- Ingest reversal events from processors (chargebacks, retrievals, pre-arbitration), ACH return files, order management (refunds, returns), claims systems and rewards ledgers into one per-entity history.
- Link each reversal to the original transaction and to the decision that approved it. Otherwise you can't attribute losses or retrain models (Loss Taxonomy and Attribution).
- Store reason codes and outcomes along with counts. A lost dispute and a won dispute mean opposite things.
- Build vs. buy: your own history is built; consortium data on dispute and return behavior across merchants, and retailer return-authorization services, are bought (Consortium and Data Sharing).
How to use it
- Exposure and eligibility: accounts with recent unauthorized ACH returns lose instant availability; serial disputers get tighter refund and claims policies; reward accounts with frequent reversals get earnings vested past return windows (Reward Vesting and Clawback).
- Classify reversals: separate merchant-error disputes (legitimate) from first-party misuse, and distinguish third-party fraud victims (whose cards were stolen) from serial disputers. FraudClassifier-style taxonomies help keep labels consistent.
- Combine with usage evidence: device, login and delivery data showing the disputing customer received value support both decisioning and representment (Device Fingerprinting, Login Anomalies).
- Link across entities: a "new" account whose instrument, device or address carries prior abuse history is not new (Entity Link Analysis).
- Returns policy: use history to differentiate returns policy (instant refund vs. refund on receipt of item) as well as risk scores. See Claims and Returns Policy Controls.
Strengths and limitations
Strengths: highest-quality labels available; strong predictor of first-party abuse; supports both prevention and dispute recovery.
Limitations: the signal lags. Card disputes can arrive months after the transaction, and consumer ACH unauthorized returns use an extended return window measured in weeks (see Nacha Rules and ACH Risk), so new actors have no history. Account-level history can be degraded by account cycling (linkage compensates). Reason codes are noisy, and customers choose the code that wins. Because reversal history is a lagging label, pair it with leading signals.
False positives and fairness
- Genuine victims of third-party fraud who dispute multiple charges after a card compromise.
- Customers who return frequently for legitimate reasons, such as sizing in apparel, gift returns and defective products.
- Customers of merchants with poor fulfillment who correctly claim INR/SNAD.
- People with disabilities or limited mobility who rely on generous returns.
- Extreme couponers whose purchases legitimately get returned when stores misapply offers.
Avoid labeling customers as fraudsters based on dispute counts alone; reason codes and outcomes matter. Customers have statutory dispute rights under Reg E and Reg Z that controls must not chill.
Privacy and compliance considerations
Reg E and Reg Z protect consumers' right to dispute; using dispute history to penalize customers must not discourage legitimate claims or violate error-resolution obligations. Consortium sharing of returns and dispute history may implicate FCRA if used for eligibility decisions, so get legal review. Network rules govern dispute data use. See Regulation E and Unauthorized Transfers, Card Network Dispute Rules and Privacy and Data Use in Fraud Prevention.
Mechanisms it helps detect
Chargeback and Friendly Fraud · Refund and Return Fraud · Buyer Protection Claim Abuse (INR and SNAD) · ACH Funding Fraud · First-Party Fraud · Card-Linked Offer Abuse · Receipt Fraud (Cashback) · Bust-Out Fraud
Related signals and controls
Customer Tenure and History · Payment Instrument Signals · Funding Source Signals · Basket and SKU Anomalies · Claims and Returns Policy Controls · Negative Lists and Deny Lists · Playbook - Chargeback Management Members · Playbook - Refund and Claims Abuse Members · Playbook - ACH Funding Fraud Members
Sources and further reading
- Qualpay: Visa Compelling Evidence 3.0: CE 3.0 effective April 2023; prior undisputed transaction criteria.
- Checkout.com: Visa CE 3.0 overview: matching data elements and 120–365 day window.
- 12 CFR 1005.11 (Regulation E error resolution): consumer dispute timelines and provisional credit.
- FedPayments Improvement: FraudClassifier model: consistent classification of authorized vs. unauthorized fraud.