Fake Orders and Chargebacks: How Do You Screen Peak Traffic Without Killing Sales?

You screen peak traffic by tightening checks on the small slice of orders that carry real loss, and leaving the rest alone. In practice that is high value orders, a delivery address that does not match the billing details, a rush delivery request, a first-time buyer using a freshly added payment method, and several attempts on the same card in a few minutes. Those patterns cover most of the money. A blanket rule that holds every order over a certain value will quietly cost you more in abandoned carts than the fraud it stops, and you will never see that number on a report.

What actually changes between now and 11.11?

Three things move at once. Order volume rises, so a fixed percentage of bad orders becomes a bigger absolute number. The share of first-time buyers rises, because discount traffic is not your regular customer base, and first-time buyers are where almost all card fraud sits. And your own tolerance drops, because during peak nobody wants to be the person who held up a shipment to make a phone call.

That last point is the real risk. Fraud controls do not usually fail because the rules were wrong. They fail because the person who was supposed to review the held orders was packing boxes, and by day three the queue was cleared without being read.

Which fraud actually costs you money?

Separate the loss types, because they need different answers.

• Stolen card on your own webstore. You ship the goods, the real cardholder disputes the charge, and you lose the goods and the payment. This is the expensive one.
• Friendly fraud. A genuine customer disputes a genuine purchase, often months later, claiming the item never arrived. Winnable, but only if you kept the delivery evidence.
• Fake payment confirmation. A buyer sends a doctored PayNow or bank transfer screenshot. PayNow is a push payment with no dispute mechanism, which cuts both ways: nobody can claw money back from you, and nobody can get your goods back either. Match payments against the bank statement, never against an image.
• Delivery refusal on cash on delivery. Not fraud in the legal sense, but you pay freight out, freight back, and handling, on an order that was never real.
• Marketplace disputes. On Shopee or Lazada the platform holds the payment and runs the dispute, so your exposure is returns, not-received claims and your seller account health score rather than a bank chargeback.

Your own checkout is where the serious money sits. Marketplace orders hurt your metrics; webstore orders hurt your bank account.

What does one chargeback really cost?

Add it up properly, because the headline amount is the smallest part. You lose the sale value. You lose the goods, which are already out the door. Your payment provider charges a dispute fee, commonly in the range of fifteen to thirty dollars per case regardless of outcome. Someone spends an hour or two assembling evidence and uploading it. And the win rate on contested disputes is modest, often well under half, so you should budget as though you will lose a good share of them.

A ninety dollar order that goes to dispute is realistically a two hundred dollar event once the fee, the stock and the staff time are counted. There is a second cost that bites harder: card schemes monitor merchants whose dispute ratios run high, and sustained levels around or above roughly one percent of transactions attract penalties, rolling reserves or, at the extreme, loss of card acceptance. Ask your acquirer for your current ratio before peak. Most SME owners have never seen it.

What can you switch on this week?

None of this needs a new system. Nearly all of it is already in your payment gateway and switched off.

• Turn on 3-D Secure authentication. When a card payment is authenticated this way, liability for fraud disputes generally shifts to the card issuer rather than you. Modern versions challenge only a minority of transactions, so the checkout friction is far lower than the old experience people remember.
• Enable the gateway's own risk scoring and set it to review, not block, for the first two weeks. You want to see what it catches before you let it reject customers.
• Set one hold rule with a value threshold that makes sense for your basket, plus the address mismatch flag. One rule that gets reviewed beats six that do not.
• Cap velocity. Multiple failed card attempts from the same session is the single clearest signal of card testing, and it costs you nothing to block.
• Require proof of delivery with a photo or signature on anything above your hold threshold. Ask your courier what they capture and how long they keep it. Many keep tracking data for a limited window, which is shorter than the period a cardholder has to dispute, and that window can run to around 120 days.

Who decides, and what do you write down?

Before volume arrives, put one page in front of the team. It needs five things: the value above which an order is held, the named person who clears the queue each morning and the backup when that person is on leave, the script for the verification call, the list of evidence to keep for every held order, and the date each week when someone checks the dispute count.

The evidence list matters more than the rules. For a contested dispute you want the delivery tracking number and proof of delivery, the authentication result from the gateway, the IP and device details from checkout, and any messages with the customer. Collected at the time, that pack takes two minutes. Reconstructed in February for a December order, it often cannot be assembled at all, and the dispute is lost by default.

One more line for that page: a rule that nobody changes a delivery address after payment without a verification call. Address change after authorisation is one of the most reliable fraud signals there is, and it is the request most likely to be waved through by a helpful person under pressure.

Frequently asked questions

If we turn on 3-D Secure, are we safe from chargebacks?

No. Authentication generally moves liability for fraud disputes to the card issuer, which removes the worst category. It does nothing for disputes raised on the grounds that the item never arrived or was not as described. Those still land on you, and they are won or lost on delivery evidence.

Should we keep offering cash on delivery through peak?

Only if you have measured your refusal rate. Work out what each refused delivery costs you in freight out, freight back and handling, then compare that to the orders you would lose by removing the option. Many sellers find cash on delivery pays for itself in normal months and loses money in December, when refusal rates climb.

What dispute rate should we be worried about?

Ask your payment provider for your own figure rather than relying on a benchmark. As a working guide, treat anything approaching one percent of transactions as a problem that needs attention now, because scheme monitoring programmes and the penalties attached to them start to apply in that region. Check it weekly during peak rather than monthly, so a bad pattern is three days old when you find it instead of five weeks.

The goal for the next three weeks is not zero fraud. It is knowing your current dispute ratio, having one hold rule that a named person actually reviews, and keeping the evidence that lets you contest the disputes worth contesting.