What Our Own COD Data Showed About Returns

The blended number tells you nothing useful

A single COD failure rate across the whole store is an average of several very different behaviours. Returning customers and branded search traffic barely fail at all. Cold prospecting from social fails at a materially higher rate, because the order was placed on impulse with no payment behind it.

Reporting one number means the campaign generating most of your losses looks the same as the one generating none of them.

Segment by channel first

Channel is the cut that usually produces the largest difference, and it is the one you can act on immediately, because COD rules can be applied differently by campaign.

A store can reasonably offer COD freely to returning customers and branded search, while requiring confirmation or restricting it above a value threshold on cold prospecting traffic. That is not a customer experience compromise. It is applying friction where the risk is.

Then segment by pincode, using your own history

Purchased pincode blocklists are built from someone else's customers buying someone else's products. Your own delivery history is a far better predictor, and after a few hundred orders you will have enough to identify the pincodes where failure is consistent rather than incidental.

The important distinction is consistency. One failed delivery in a pincode is noise. Repeated failures across different orders and different months is a pattern worth acting on.

What actually moved the number

Each of these moves the rate by a few percentage points. None of them is dramatic alone. Together they are the difference between COD being survivable and not.

• A prepaid incentive, framed as a saving rather than a COD penalty
• Confirmation before dispatch, with WhatsApp getting noticeably better response than SMS
• An order value ceiling on COD, since high-value failures hurt most
• Address quality validation at checkout, which catches a share of undeliverable orders before they ship