Anti-Fraud

A customer can carry flags: several addresses on one device, several on one network, a high refund rate, a lost dispute, or payments spread across many providers. Each flag is a separate fact with a sentence explaining it, shown to you on the customer and order. Flags are never added up into a score, and on its own a flag never blocks a checkout, holds an order or refunds anything. You decide what to do, and you can decide in advance to hold flagged orders until you have looked at them (below).

Anti-Fraud → Blacklist. Add an email address, a network (a typed IP address or one taken from an order) or a device (taken from an order, since nobody has a raw device fingerprint to type). Checkout refuses a matching customer with a clear message, and the entry counts its hits. Removing an entry takes effect on the next attempt.

The blacklist is not checked when an accepted quote, a free lead-magnet claim or a licence change creates an order, because those do not go through checkout.

Anti-Fraud → Whitelist silences flags about the identifier it names and nothing else. Whitelisting an email hides that email’s own flags, but not the flags on its device, and never overrides a blacklist entry on a different identifier. One identifier cannot be on both lists.

Anti-Fraud → Fraud logs lists the flags on customers who ordered in a date range, worked out fresh when you open it. It covers the 200 most recent customers in the range.

Every paid order has a Was this order fraud? card with two buttons: Mark as fraud and Mark as legitimate. Click again to change your answer, or Clear to take it back. Your answer changes nothing on the order. There is no refund, block or blacklist entry. Sellstamp counts it, with refunds and chargebacks, to check how often each flag turns out to be right, and only ever as totals across all stores.

Fifteen checkouts from one browser, or fifty from one network, at your store within ten minutes, and that browser or network is blocked for five minutes. The block doubles on each repeat, up to an hour. The person sees how long to wait. Your store’s status is never touched.

Under Store settings → Policies you can hold delivery of orders over an amount, optionally only from devices that have not paid at your store in the last 90 days. The customer is still charged. The download, key or parcel waits until you approve it.

You can also hold an order a fraud flag fired on: any flag, or only a serious one such as a lost chargeback. That hold releases itself after the number of hours you choose, so a sale is not lost while you sleep, and the buyer sees when that will be. Approve delivers at once; cancel refunds in full. The order’s page lists everything that happened automatically.

Under Anti-Fraud → Rules you can write rules such as “if the order total is at least 500 and it is the buyer’s first order, hold delivery”. A rule acts at checkout (refuse with a reason the buyer reads, ask for a CAPTCHA, or ask the card’s bank for 3-D Secure on Stripe) or once paid (flag, hold, or wait for the buyer to confirm their email before the download is released; the receipt is sent either way). Try a rule on your past orders before switching it on: you see how many of your last 300 orders it would have matched.

The device and network are stored only as keyed hashes, never raw, and are kept for 90 days, beside the country of the buyer’s time zone and a rough checkout speed (never the time zone or the exact timing). A record of which flags fired on each paid order, and your fraud or legitimate answer, holds no hash or address and is kept for 365 days. When your payment provider reports a payment paid, what it said about that payment (card checks, 3-D Secure, the card’s country and type, earlier declines) is kept for 90 days, with the card only as a keyed hash. There is no geolocation and no third-party fraud service. You can stop collecting them under Store settings → Policies, but that also switches off the card-testing block, which works by counting exactly these records.