Chargebacks: the ratio decides your cost, not the disputed amount
Scheme monitoring works on a monthly ratio whose numerator and denominator describe different cohorts. In a falling month the ratio rises while disputes fall, and nothing about your book has changed.
The money lost to a chargeback is the smallest part of what a chargeback costs. The expensive part is a ratio computed by the card scheme, against a denominator you do not control, which decides whether your acquirer keeps pricing you as a normal merchant. Most brokers manage the disputes and never look at the ratio until an acquirer raises it.
The ratio, and the timing trap inside it
Card scheme monitoring programmes work on a monthly dispute ratio combined with an absolute count. Both tests have to be understood together, because a small merchant can breach on ratio while a large one breaches on count, and the remedies are different.
ratio = disputes_received(month M) / transactions(denominator month)A dispute arrives weeks or months after the transaction that caused it. When the numerator counts disputes received this month and the denominator counts transactions processed this month, the two are measuring different cohorts. In a stable month nobody notices. In a month where deposit volume falls, the numerator is still being fed by the large months behind it while the denominator shrinks, and the ratio rises with no change in client behaviour at all.
A broker can improve its fraud controls, reduce disputes in absolute terms, and watch its monitored ratio go up, because deposits fell faster than disputes did.
The lifecycle decides what you can defend
A dispute is not one event. It is a sequence with defined windows, and the defence available at each stage is different. Treating the whole thing as one queue is why representment win rates stay low.
- 01Retrieval or inquiryissuer wants context, no funds moved
- 02Chargeback raisedfunds debited, reason code assigned
- 03Representmentyou submit evidence against that code
- 04Pre-arbitrationissuer rejects, narrower question
- 05Arbitrationscheme decides, filing fee at risk
Reason codes separate into two families that have almost nothing in common. Fraud codes assert the cardholder did not authorise the transaction, and are answered with authentication evidence: 3-D Secure results, device and IP consistency, prior undisputed deposits from the same card. Service codes assert the cardholder did authorise it but did not receive what was promised, and are answered with the account record: the deposit crediting to tradeable balance, the trading activity that followed, the terms accepted at signup.
- Submitting trading history against a fraud code answers a question nobody asked. It loses.
- Submitting an authentication log against a service code concedes the point. It also loses.
- A liability shift from successful 3-D Secure authentication removes most fraud codes before they are raised, which is worth more than any improvement in win rate.
Representment is an expected value calculation
Fighting every dispute is as wrong as fighting none. Each representment costs analyst time and, past pre-arbitration, a filing fee that is lost whichever way the decision goes. The decision rule is arithmetic.
EV = p(win) × (amount + fee_refund) − cost_to_assemble − p(lose) × fee_at_riskThe second term is the one brokers underestimate, because assembling evidence for a CFD deposit dispute means pulling the authentication record from the PSP, the crediting entry from the ledger, the trading activity from the platform and the accepted terms from the onboarding system. Four systems, one deadline, per dispute.
| Fight everything | Fight by code and evidence | |
|---|---|---|
| Disputes contested | 400 | 164 |
| Analyst hours | 340 | 148 |
| Win rate | 21% | 58% |
| Amount recovered | $71,400 | $96,300 |
| Arbitration fees lost | $9,600 | $2,100 |
| Net of analyst cost at $45/h | $46,500 | $87,540 |
Illustrative figures, chosen to show the mechanism. Your own numbers will differ.
The cheapest dispute is the one that becomes a refund
A meaningful share of service-code disputes begin as a client who could not withdraw, could not reach support, or did not understand a charge. Those clients call the issuer because calling you did not work. Every one of them is a dispute you could have converted into a refund at a fraction of the cost, and refunds do not enter the monitored ratio.
The signals that precede a dispute
- A withdrawal request open beyond your own stated service level, on an account that deposited by card within the chargeback window.
- A support ticket containing dispute language, unresolved, on a funded account. This is the highest-yield signal and it sits in a system nobody joins to payments.
- A first deposit from a card with no subsequent trading activity, which is the standard shape of both genuine fraud and genuine regret.
What to run continuously
Three series, none of which any single system can produce on its own.
The ratio, computed the way the scheme computes it
Not your internal dispute rate. The monitored ratio, with the denominator month the programme actually uses, projected forward on current volume so you see a breach coming with a month to act rather than learning about it from your acquirer.
Win rate by reason code, not blended
A blended representment win rate is an average across two populations with different defences and different achievable outcomes. It cannot be improved because it does not describe anything you can change.
Time to evidence
How long it takes from a dispute landing to having all four pieces of evidence in one place. If that number is close to the response deadline, the win rate is being set by administration rather than by the merits, and no amount of arguing better will move it.