Fraud Rules That Do Not Cost You Sales
Blunt fraud filters reduce fraud and revenue at the same rate. Segmented rules and a review queue do better than a global threshold.
Most disputes are operational failures wearing a payments costume. Sort them by cause and the fixes are cheaper than you expect.
Before changing anything, pull ninety days of disputes and sort them by reason. The distribution almost always concentrates in two or three causes, and each needs a different remedy.
Buying a fraud-screening tool to solve a subscription-cancellation problem is a common and expensive mistake. The tool works; it is aimed at the wrong target.
The cheapest chargeback reduction available to most merchants is a recognisable billing descriptor. A customer scanning a statement for something they recognise will dispute what they do not.
Include a brand name the customer actually saw at checkout, plus a support contact where the format allows. Then confirm what appears on a real statement rather than trusting the configuration screen — descriptors are truncated and reformatted in ways that are not obvious until you look.
Disputes for non-receipt cluster in the gap between purchase and arrival. Silence in that gap invites the customer to conclude something has gone wrong.
A customer who cannot reach you within a day or two will call their bank instead. From their perspective the bank is simply a support channel with better response times.
Publish a contact route that is monitored, answer refund requests quickly, and process approved refunds promptly. A refund is expensive; a chargeback is more expensive and counts against your ratio.
For recurring billing, cancellation friction is the single largest source of preventable disputes. If a customer signed up in two clicks and cancelling requires an email and a three-day wait, some of them will dispute instead.
Offer self-service cancellation through the same channel used for signup, confirm it in writing, and stop billing immediately. Retention flows are legitimate; a retention maze is a dispute generator.
When a dispute does arrive, a representment is an evidence exercise on a deadline. Build one template per reason code you actually receive, so nobody writes a response from scratch at 6pm on the due date.
A strong pack usually includes the transaction record, evidence of consent, proof of delivery or access, the support history, and the policy the customer accepted. Outcomes are decided by the issuer and the network, and no representment can be guaranteed.
Card networks operate monitoring programmes with defined thresholds, and processors calculate ratios in ways that may differ from your internal reporting — count-based or value-based, against current-month or prior-month volume.
Ask your provider how yours is calculated, then track that number. Being surprised by a monitoring notice is almost always a measurement failure rather than a sudden change in customer behaviour.
Educational content only
Educational content only. Nothing here is legal, tax or financial advice, and none of it guarantees an outcome with any provider.
Have a question this article did not answer?
Describe your setup and we will answer the specific question — including when the honest answer is that a route is not realistic.
Keep reading
Adjacent topics that tend to matter for the same decisions.
Blunt fraud filters reduce fraud and revenue at the same rate. Segmented rules and a review queue do better than a global threshold.
Recurring revenue is underwritten on three things: whether the customer knew, whether cancelling was easy, and whether you can prove both.
Tell us about your business, current processing situation, and growth plans. Our team will review the information and explain the available next steps.
Submitting a form does not guarantee approval and does not create a contractual relationship. It begins a review conversation so we can explain the options that may be available to your business.