Cutting chargebacks in your first month

Chargebacks cost you the goods, the payment, and a fee on top — and enough of them will put your payment account at risk. Here is a practical order of work for a store that has started seeing them, roughly cheapest and fastest first.

1. Find out what kind of chargeback you have

Before changing anything, look at your last few disputes in Stripe or PayPal and read the reason codes. They fall into two very different groups:

• Genuine fraud — a stolen card was used on your store. The real cardholder disputes a purchase they never made.
• “Friendly” fraud and service disputes — a real customer really did order, but does not recognise the charge, expected a different delivery date, or asked for a refund and did not get one.

These need opposite fixes. Fraud screening does nothing about a customer who did not recognise your billing descriptor. Sorting your disputes into these two buckets is fifteen minutes of work and it decides everything you do next.

2. Fix the free things first

A surprising share of disputes are self-inflicted and cost nothing to fix:

• Your billing descriptor. If your store is “Ortica Goods” but the card statement says “OG-COMMERCE LTD”, some customers will dispute a charge they simply do not recognise. Set it to your trading name in your payment provider’s settings.
• Order confirmation emails. They should arrive immediately, state the trading name, and say when delivery is expected.
• A findable refund policy and a reachable support address. A customer who can get a refund from you has no reason to ask their bank instead. Disputes are frequently a support failure wearing a fraud costume.
• Delivery expectations. “Ships in 2–3 weeks” on the product page prevents a dispute that “where is my order?” would otherwise become.

3. Turn on screening — in review mode

Now add fraud screening — this is what WooFraudGuard is for — and resist the temptation to have it cancel orders on day one. Run it in review-only mode for two weeks and just watch. You are looking for two things: which signals actually fire on your traffic, and whether anything legitimate is getting caught.

Every store has its own normal. A store selling gifts sees a lot of legitimate billing-and-shipping mismatches. A store selling software sees almost none, so a mismatch there means much more. Thresholds copied from someone else’s store are a guess; thresholds set from two weeks of your own orders are evidence.

4. Tighten from what you saw

After a couple of weeks, set your review threshold where it catches the patterns you saw on real problem orders, and set the cancel threshold well above it. Cancelling should be reserved for orders you are confident about, because a false positive there is a lost sale and an annoyed customer.

Weight the individual rules to match what you learned. If disposable email domains never once appeared on a real problem order, that rule should not be pushing orders into review on your store.

5. Actually fight the disputes you should win

Many merchants never respond to disputes, which is why the numbers look so grim. If you have delivery confirmation, an IP that matches the billing region, and a clean order history for that customer, you have a case worth submitting. Collect that evidence at order time, not when the dispute lands 60 days later.

6. Feed the outcome back

Whatever tool you use, make sure dispute outcomes get recorded against the original orders — WooFraudGuard does this automatically through its Stripe and PayPal dispute sync. Screening that never learns which of its decisions were right is screening frozen on the day you configured it. Reviewing that once a quarter is enough to keep your weights honest.

A realistic expectation

Steps 1 and 2 often move the number more than anything else, and they are free. Screening prevents the fraud category but cannot prevent a customer disputing a charge they genuinely made. Anyone promising to eliminate chargebacks is selling something — the goal is to get them down to a level your margins can absorb, and to stop the ones that were never really fraud in the first place.