Payout held · account review
Support does not freeze a payout because of one angry email. They freeze it because a ratio crossed a line, and the ratio is arithmetic: disputes divided by sales over a rolling window. That number is recoverable — case by case — before you write the appeal.
One payment of USD 32.55. No subscription. Tax, if any, is calculated by Gumroad at checkout for your billing country.
Two different events get called "frozen", and they are fixed differently. Read the notice before you write anything.
| What the notice says | What is really being measured | What moves it |
|---|---|---|
| Payout on hold / release date pushed | Risk review of the account, usually after a dispute cluster or a first large month | Clean dispute record, delivered product, responsive replies |
| Account under review / selling paused | Dispute rate, refund rate, and complaint pattern at once | Lowering the rate by winning or preventing disputes, not by arguing |
| Reserve held / funds rolled back | Exposure on orders still inside the dispute window | Time plus evidence that the disputed orders were delivered |
Platform thresholds are not published, and anyone quoting you an exact internal number is guessing. What is published is the card network side: Visa's dispute monitoring programme flags a merchant at 0.9% of transactions in a monthly cycle and escalates at 1.8%. Mastercard runs a separate programme with its own counts. Those are the numbers acquirers and processors watch, and a processor that sees you cross them will act before the network does.
0.9% is one dispute in 111 sales. A store doing 40 orders a month goes over on a single chargeback. That is why "I only had one dispute" and "you are over the line" are both true at the same time.
An appeal that says "please review" gets a template reply. An appeal that says "here are 6 disputes, 4 were lost on missing proof, 2 are still open, here is the proof set for each code" gets read by a human. The difference is whether you can name the reason code for every dispute in the window.
The kit does exactly this loop on your machine: enter the code, tick what you actually hold, get the completeness score, the missing-item list, and the rebuttal letter with exhibits numbered to match your claims.
Published price lists from public dispute-service pages, read 2026-11-06. Same deliverable: a reason-code-matched evidence pack and a rebuttal letter.
| Option | Price | What you actually get |
|---|---|---|
| Dispute agency — complex case | $2,500 | A person writes the pack for you |
| Dispute agency — held-funds release | $1,000 + 15% | Of the funds being held |
| Dispute agency — evidence-backed appeal | $399 | Document pack assembled by a person |
| Dispute agency — basic appeal | $199 | Template appeal, one submission |
| Chargeback Reply Kit | $32.55 | You run it: scores the case, lists missing evidence, writes the numbered letter. Unlimited cases. |
One $399 appeal costs 12x this kit. If you handle more than one dispute a year, the second one is free.
Every product page on the store shows its own test output, including the negative control. Re-run it after download.
$ python3 chargeback_kit.py --selftest
[ok] visa_10_4 evidence set loaded
[ok] visa_13_1 evidence set loaded
[ok] mc_4837 evidence set loaded
[ok] amex_c08 evidence set loaded
[ok] score_partial case scores partial, lists the gaps
[ok] letter_render exhibits numbered to match claims
[ok] unknown_code fails closed, does not guess <- negative controlThe negative control is the point. An unknown reason code must refuse to invent an evidence set. A tool that guesses is worse than no tool, because you file a wrong pack with full confidence.
No account, no API key, no telemetry. Nothing you type leaves the machine.
No tool can promise that, and anyone who does is selling you hope. What it fixes is the mechanical failure: appeals get refused when the pack answers a different question than the code asked, or when an item is simply missing. Fix that first, then appeal.
No. It is a document assembly tool. It matches your case to the published evidence requirement for the reason code and writes the letter. You file it yourself.
Send the error output. We fix it until it runs, or you get a re-delivery at no cost within 14 days.
No. It runs offline on Python 3. There is no upload, no telemetry, no account.
Not sure this is your problem yet? Take the free 15-check store ops sheet first. If an open dispute is not on the list this week, you do not need the kit yet.
Get the free sheetCheckout is handled by Gumroad — card and PayPal accepted, tax included in the shown price. Product status verified live: product=gkdkb, archived=false, price 3255 USD cents.