Card Testing Bots: Small Charges That Mean a Bigger Attack
A run of one dollar authorisations is not a customer. It is a bot testing stolen card numbers, and the store that ignores it will see the chargebacks next.
A store's payment logs show fifty one dollar authorisations in an hour, all declined, all from different card numbers, all through the guest checkout. Nobody complains, because no customer was charged. The store ignores it, and two weeks later the chargebacks arrive, because the bot that tested those cards found the ones that worked and used them elsewhere, and the store's payment processor now has a fraud problem with the store's name on it.
Card testing is the quietest bot attack in ecommerce, and the most expensive in the long run. The charges are small, the damage is not, and the store that treats a run of small declines as noise is the store that gets hit next.
How card testing works
A card tester takes a list of stolen card numbers, often bought in bulk after a breach, and runs them through a store's checkout in small amounts, usually a dollar or less, to find which ones are still valid. The store's checkout is the test instrument, because it returns a clear signal: approved or declined. The bot does not need to buy anything. It needs the answer, and the store gives it for free.
The pattern is recognisable. Many attempts in a short window, small identical amounts, different card numbers with the same billing details, guest checkout, and a high decline rate concentrated on one payment method. The bots use residential proxies to spread the attempts across IP addresses, so a simple rate limit on one address does not stop them.
Why the store pays for it
The store pays in three ways. First, the payment processor charges a fee for every authorisation attempt, approved or not, so a thousand tests cost real money. Second, the approved cards get used for fraud elsewhere, and the chargebacks land on the merchants who accepted them, not just the store that tested them. Third, the processor watches the decline rate, and a store with a sustained high decline rate risks higher fees, a reserve, or an account review.
The reputational cost is quieter. A customer whose card was tested on your checkout will see the authorisation on her statement, and she will associate the store with the fraud, even though the store was the victim. The store did not steal the card. It lent the thief its checkout to test it.
Stopping the test without stopping customers
The effective controls are the ones that raise the cost of testing without raising the cost of buying. A CAPTCHA on the payment step after a small number of failed attempts, velocity limits on declines per card fingerprint, and a minimum order value that makes testing uneconomic, all help. Address verification and CVV checks should be enforced, because a tester with only a card number and no CVV will fail them.
Monitor the decline rate by payment method, not just in aggregate. A sudden rise in declines on one card brand, or a run of identical small amounts, is the signal. Alert on it the way you would alert on a checkout outage, because by the time the chargebacks arrive, the testing is long over and the damage is done.
Related reading: Glossary, About, Bots at the API Layer: Probing Product, Search, and Stock Endpoints.
Frequently asked questions
What is card testing?
A bot runs stolen card numbers through a store's checkout in small amounts, usually a dollar or less, to find which ones are still valid. The checkout returns an approved or declined signal, and the bot uses the approved cards for fraud elsewhere.
How do I recognise card testing on my store?
Many attempts in a short window, small identical amounts, different card numbers with the same billing details, guest checkout, and a high decline rate concentrated on one payment method. Residential proxies spread the attempts, so a single-IP rate limit will not stop it.
Why does card testing cost the store money?
The processor charges a fee for every authorisation attempt, approved or not. Approved cards get used for fraud elsewhere, generating chargebacks. A sustained high decline rate can also bring higher fees, a reserve, or an account review from the processor.
How do I stop card testing without blocking real customers?
Use a CAPTCHA on the payment step after a few failed attempts, velocity limits on declines per card fingerprint, and a minimum order value that makes testing uneconomic. Enforce address verification and CVV checks, and alert on a sudden rise in declines by payment method.