A delivery agent is standing outside a flat in Andheri with your box. The whole thing takes forty seconds. The customer looks at the parcel, says they don’t want it any more, and shuts the door.
Nothing unusual happened there. No fraud, no dispute, nobody lied to anybody. The customer did exactly what you invited them to do at checkout, which was to decide later.
Because that is what cash on delivery actually is. Not a payment method. An option, in close to the financial sense of the word: the customer gets the right to complete the purchase without the obligation to, exercisable on sight of the goods, at no cost to them.
You sold that option for nothing. Your return-to-origin line is the premium you forgot to charge.
Which changes the question. Stop asking how to get rid of COD. Start asking what the option costs you, who should get one, and what you should charge the people who want it.
“COD fraud” is mostly the wrong word
I’d argue with the vocabulary first, because it is doing real damage to how brands spend money on this.
Some of it genuinely is fraud. Fake orders placed to a wrong address, competitor mischief, somebody’s number used without permission. Real, and a minority.
The rest is a customer exercising a term you wrote. They changed their mind. They found it cheaper on a marketplace three days later. They ordered two sizes meaning to keep one. The parcel turned up on a bad day and saying no was the easiest thing available.
That is not fraud. That is the option behaving exactly as designed.
And the word decides where you spend. If the problem is fraud, the fix is detection, blocking, somebody’s risk engine, a line item. If the problem is an option you priced at zero, the fix is in your own checkout, and you can start on Monday without buying anything.
What the data can and cannot carry
Here is the one source I’d trust to set the scale, and then the honest limits on it.
Unicommerce built its FY2026 India D2C report from its own platform data, 410 million-plus shipments across 6,000-plus brands. Their published figures put overall RTO at 39.2% in November 2025, falling to 25.6% by January and 21.0% in March 2026. For the festive quarter they split it by payment mode: COD orders returned at 58%, prepaid under 15%.
Read those carefully. The 58% is a festive-quarter peak, not an annual average, and the all-order numbers aren’t directly comparable with the mode-split ones. It’s also a vendor’s data about brands running that vendor’s software. Platform data from 410 million shipments is still the best thing available here, which tells you how little independent measurement exists.
The gap between modes is the part nobody disputes. Every source I could find puts COD RTO at a large multiple of the prepaid rate. That ratio is the thing to reason from, not any single headline percentage.
Now the figure I will not give you. I went looking for what share of Indian ecommerce orders are COD and found published claims ranging from under a fifth to nearly two thirds, depending on whether the source counts orders or payment value, D2C or all ecommerce, and which year. Lining those up produces a trend that is not there. So don’t plan around a number you read, mine included. Pull your own split, by category and by pin code.
Price the option from your own courier invoice
Twenty minutes, and it needs nothing but documents you already have.
Open last month’s courier invoice and your ad account, and write down four numbers for one category.
Forward freight per shipment. Reverse freight on a failed delivery, usually close to the forward leg rather than free. Handling on the parcel that comes back: receiving, inspection, repack, the units that arrive unsellable. And your acquisition cost on that order, the one people leave out, because the money you spent on Meta to win it is gone whether the box came back or not.
Add those four. Multiply by your COD RTO rate for that category. That product is the expected cost sitting on every single COD order you accept, including the ones that deliver perfectly.
Then subtract it from the contribution margin on a COD order and look at what is left.
The line that does the damage is almost never the freight. Freight is tens of rupees a leg. Acquisition is hundreds, which is why leaving it out of the sum makes COD look survivable when it isn’t.
A note on that, since it’s contestable. Some operators exclude CAC on the grounds that it’s already sunk at the point of dispatch. Mathematically fair. But if the question is whether to offer the option at all, the sunk cost is precisely what the option destroys, so it belongs in the sum. Exclude it when you’re negotiating freight rates, include it when you’re designing checkout.
Everything that works does the same single thing
Look at the levers that reliably help and they have one mechanism in common. They move the decision point back from the doorstep to the checkout, where you are still in the conversation.
Partial advance is the cleanest of them, and the one I’d reach for first, because it does not ban the option, it prices it. Rs 100 or Rs 200 taken upfront, the rest on delivery. The customer who intended to buy barely notices. The customer who wanted a free look at the product now has something to lose by walking away, which was the entire missing ingredient.
A prepaid incentive is the same move from the other side, and it is worth more than the discount costs, because you are not buying a conversion, you are buying certainty and your working capital back. Which matters more than founders expect, since COD money sits with the courier until remittance and a brand growing fast on cash orders can be profitable on paper and short in the bank.
Address and phone confirmation before dispatch helps for a duller reason: a meaningful share of failures aren’t refusals at all, they’re a wrong flat number or a phone nobody answers.
And holding first-time buyers to prepaid in your worst pin codes is the narrowest version, which is why I like it. Not a blanket rule, a rule aimed at the specific intersection where your own data says the option gets exercised.
What about collecting digitally at the door, agent with a QR code instead of change? It solves the cash handling and the remittance lag, both real. On whether it cuts refusals I’d be careful: the claims I could find come from the companies selling it, and the mechanism argues the other way, because the customer still decides at the door and the door is the problem.
Where this ends up
The brands that fix this will not be the ones that switched COD off. That is the expensive answer, and it fails on selection: the orders you lose first are the ones that would have delivered cleanly, because the customer who genuinely prefers paying cash is also the customer who answers the phone and takes the parcel. You keep the option-takers and lose the buyers, which is backwards.
The brands that fix it will be the ones that stopped treating it as free and universal. Priced for first-time buyers, open for repeat ones, tuned by pin code and category, with an advance on the orders where the data says the option gets used.
My actual prediction, and you can hold me to it. In two or three years a free, unconditional COD option on every order in every pin code will read the way unlimited free returns in apparel read now: a tactic from a period when somebody else was funding growth, kept long past the point where the arithmetic worked.
If you’d like a second pair of eyes on what your COD mix is really costing you by category, drop me a line on email, WhatsApp, phone or LinkedIn and we can have a quick chat, and my notes on performance marketing cover the acquisition side of the same sum. I’m taking on a small number of contracts at the moment, so I’ll tell you straight whether it’s something I could help with.