Building a chargeback dispute playbook (with templates)
- Win rates by reason code vary wildly — code 4853 (cancelled subscription) wins 60-70% with the right evidence; code 4837 (no auth) wins under 10%.
- Every operator should keep five evidence templates pre-built: tracking, signed agreement, login history, IP+device match, and prior dispute history.
- Representment is a numbers game — file every dispute, including the ones you think you'll lose. The marginal cost is tiny; the cumulative win rate compounds.
On this page
Most operators treat chargebacks as a tax. They shouldn't. With a real playbook — pre-built evidence templates, reason-code routing, and a queue you actually work through — you'll win 35–55% of disputes that you currently lose by default. On a portfolio doing $2M/year with a 1.0% chargeback ratio, that's $7k–$11k of recovered revenue per year and (more importantly) the difference between staying under VAMP threshold and getting frozen.
This is the operator's playbook. We'll cover the win rates by reason code, the five evidence templates every brand should pre-build, the workflow for routing disputes by type, and the templates themselves.
Step 1: Know your win rate by reason code
Not all chargebacks are equal. Win rates vary by reason code based on what evidence the issuing bank will actually credit. Rough industry benchmarks for CNP ecommerce:
- 4853 (cancelled recurring transaction): 60–70% with subscription log + cancellation policy + evidence customer used product after charge.
- 4837 (no cardholder authorization): 5–12%. The issuer almost always sides with the cardholder. Only winnable with EMV-3DS authentication on file.
- 4855 (goods/services not received): 55–75% with delivery confirmation + signature.
- 4863 (cardholder doesn't recognize): 65–80% with clear billing descriptor + receipt + IP/device match.
- 13.1 (Visa: services not provided): 40–50% with proof of service delivery + customer login records.
- 10.4 (Visa: other fraud — card-absent environment): 8–15% — same as Mastercard 4837. Hard to win.
- 13.6 (Visa: credit not processed): 50–60% — usually a refund timing dispute. Show refund issued.
The takeaway: code 4837/10.4 is rarely worth heavy evidence. Codes 4853, 4855, 4863, 13.1, 13.6 reward effort. Build your queue around the high-leverage ones.
Step 2: Build the five evidence templates
Every dispute needs evidence. Pre-built templates make submission a 4-minute task instead of a 40-minute one.
Template 1: Order proof packet
- Order confirmation email (timestamp, customer email, items, amounts)
- Stripe / processor receipt with transaction ID + descriptor + amount + date
- Shipping label generation timestamp
- Tracking number with carrier-confirmed delivery (USPS/UPS/FedEx tracking export)
- Delivery confirmation screenshot from carrier
- If signed delivery: signature image
Use for: 4855 (not received), 4863 (not recognized), 13.1 (not provided).
Template 2: Subscription / recurring billing packet
- Original signup confirmation with terms acceptance checkbox screenshot
- Welcome email with subscription terms + cancellation instructions
- Each renewal notification email with date/time
- Customer login history showing post-charge usage
- Cancellation policy URL + Wayback snapshot of policy on date of signup
- Customer service contact log (any tickets, refund offers, response timestamps)
Use for: 4853 (cancelled recurring), 13.6 (credit not processed).
Template 3: Authentication packet
- EMV-3DS authentication result (CAVV value)
- AVS match result (full match / address match / zip match)
- CVV match result
- IP address + geolocation at order time
- Device fingerprint hash
- Prior successful order history from same card / device / IP
Use for: 4837 (no authorization), 10.4 (other fraud).
Template 4: Customer communication packet
- All emails to/from customer email address
- Support ticket transcripts
- Refund offers made + customer response
- Any social media comments or reviews left by the customer
- Phone log if applicable
Use for: any dispute where customer engagement after charge demonstrates they received and used the product.
Template 5: Prior history packet
- List of prior successful orders by this customer (card hash, email, ship address)
- List of prior disputes by this customer if any (red flag for friendly fraud)
- Customer lifetime value
- Frequency of orders
Use for: friendly fraud patterns. Particularly powerful with code 4863 where "I don't recognize" is contradicted by a 12-month order history.
Step 3: The dispute workflow
Day 0: dispute hits the dashboard
Triage in <24 hours. Tag with reason code and customer email. Pull the order. If reason code is 4837/10.4 and you don't have EMV-3DS, decline to fight (acceptance pre-filed) — save the time.
Day 1–3: evidence assembly
Pull the matching template. Fill it from order management system + email + carrier API. Don't over-write a cover letter; one tight paragraph stating the case + the evidence packet attached is what the issuer reviewer reads.
Day 3–5: submit
Submit through Stripe / Authorize.net dashboard. Keep a local copy of everything submitted. Set a calendar reminder for the response deadline (usually 30–60 days from dispute date).
Day 30–60: outcome
Issuer rules. If you win, funds restored. If you lose, you can pre-arbitrate — but only if the new evidence is materially different. Pre-arbitration loses incur extra fees ($250–$500); pick battles.
The cover letter template (use verbatim, edit the brackets)
Dispute response — [Transaction ID]
This dispute (reason code [4853 / 4855 / 4863 / etc.]) is contested. Evidence attached demonstrates:
1. The cardholder placed the order on [date] from IP [x] matching billing zip [y].
2. Order was [delivered with tracking confirmation / fulfilled with login activity / charged per signed subscription terms].
3. Customer [used product on date / received delivery on date / had prior successful orders].
4. [If applicable: customer was offered refund on date and declined, see attachment 4].
We respectfully request the dispute be resolved in our favor.
Attachments: [list attachments by name]
Notes: keep it tight. Issuer reviewers spend ~90 seconds on each case. Bullet evidence, no narrative.
The metrics that prove the playbook works
- Net dispute win rate: total wins / total disputes (excluding pre-filed acceptances). Target: >40% across all codes. Operators with no playbook hit 8–15%.
- Win rate by code: track separately. Identifies which codes need template improvement.
- Time-to-submit: median hours between dispute notification and evidence submission. Target: <72 hours. Past 7 days, win rates drop sharply.
- Recovered revenue per month: wins × average ticket. Justifies the headcount or tooling spend.
- Chargeback ratio impact: wins reduce numerator on most network calculations (ratio is calculated on net disputes, not gross). Watch this against VAMP threshold.
Tooling vs DIY
For operators with >30 disputes/month: Chargeflow, Justt, or Midigator handle representment as managed service for 20–30% of recovered revenue. They're worth it if your team won't actually work the queue. They're overpriced if you have a half-time ops person who will. The math: if you win 40% of disputes at average ticket $80 with 30 disputes/month, that's $960/month recovered. A managed service takes $200–$300 of that. A spreadsheet + the templates above takes 4 hours/month of ops time.
multiflow consolidates dispute queues across every brand in the portfolio into one queue with per-reason-code routing built in. See how dispute ratios are calculated across sub-brands or the 2026 chargeback ratio guide for the network-threshold context.
The two mistakes operators make most
- Filing nothing. 60% of disputes are not contested at all. The acquirer interprets that as agreement with the cardholder. File something — even a thin packet — on every winnable code.
- Over-writing the cover letter. Three paragraphs of narrative reduce win rate. The reviewer wants evidence in 90 seconds, not a story. Stick to the bullet template above.
Build the playbook once, run it weekly, watch the recovered revenue land back in the bank. It's not glamorous; it's 4 hours a month and a real margin lever.