Fake orders and COD fraud

Not every failed COD order was ever meant to succeed. A share of COD orders are placed with no intention of accepting delivery: pranks, competitor mischief, fraud rings probing defences, bots. GoKwik reported in February 2024 that, over the two weeks it had just observed, 2 to 10% of total orders were fraudulent on affected merchants in its network, rising to 15 to 20% for some. That is a point-in-time observation during a fraud wave, not a steady-state rate, but it shows how high the spikes run. That is on top of the ordinary low-intent RTO problem.

Fraud leaves fingerprints. The signals that come up again and again across providers:

Gibberish details. Names and addresses that no human would type about themselves: random character strings, keyboard mashes, joke names.

Dead contact points. Phone numbers that are switched off, do not exist, or have never been reachable. A number that has never answered any call from any courier is not a customer.

Velocity. The same phone number, device or IP address placing many COD orders in a short window, often across multiple brands. Genuine customers do not order from six stores in an hour.

Pincode clustering. Fraud rings operate from locations. Spikes of orders from pincodes you do not usually serve deserve a second look.

Repeat refusal history. A customer who has refused three of their last five orders is a bad bet on the sixth, whether you call it fraud or habit. Without shared intelligence across sellers, each brand sees only its own slice of that history, which is why network-level tools catch what single-brand dashboards cannot.

Free: pattern checks at checkout. Phone-format validation, a landmark field, blocking obviously fake strings. Crude but catches the laziest share.

Cheap: confirmation before dispatch. Fake orders fail confirmation almost by definition; there is nobody real to confirm. OTP verification at checkout is the stronger version: GoKwik reported its fake-order catch rate rising from about 80% to 95% when OTP was added to COD orders.

Pricier: risk scoring. Checkout platforms score each order against network history: has this number ordered and accepted before, refused before, appeared across many stores? High-risk orders get held for verification or pushed to prepaid only.

The right response depends on your exposure. If fake orders are 1% of your volume, confirmation alone handles them. If you are a brand with a viral moment, a new launch or a visible competitor fight, spikes of 10% or worse are documented, and stronger gates earn their cost quickly. Watch your own weekly data for the signature: a sudden RTO rise concentrated in new customers, specific pincodes or specific products.