A cash-on-delivery post-mortem

Most discussion of cash on delivery in India is a discussion about RTO — the parcel that goes out, gets refused, and comes back. NAPPEE, an Indian baby diaper brand, had that problem too. But their data also shows a second loss, earlier in the funnel, that no RTO report would ever surface.

Across a 46-day window in which they offered COD, 178 of 415 COD orders — 42.9% — were never confirmed and therefore never dispatched. They generated work, and in some cases fees, without ever generating a sale or appearing in an RTO statistic. Roughly 15–20% of those customers later came back and bought anyway, mostly prepaid, so this is not 178 customers lost for good. It is 178 orders that cost money and attention before anyone knew whether they were real.

NAPPEE shared their operational data and their experience with CentraTalk for this research. We have not used or tested their products, and nothing here is a review of them. We received no payment for writing it and NAPPEE is not a customer.

We went looking for this because we kept hearing the same received wisdom — that COD's cost is RTO — and wanted to see whether a brand with actual numbers agreed. NAPPEE was unusual in two ways that made them worth asking: they had run COD inside a clearly bounded window, and they had kept the numbers.

The window

NAPPEE sells premium baby diapers online, from preemie and newborn sizes up to XXXL, in pant-style and tape-and-velcro formats. Their first COD transactions in the period they reviewed were on 15 November 2024. They discontinued COD around 30 December 2024. In those 46 days they took 609 orders.

609 orders in 46 days · what became of them

Cash on delivery415 of 609 · 68.1%

Prepaid194 · 31.9%

Inside the 415 COD orders

Confirmed, then dispatched237 · 57.1%

Never confirmed, never dispatched178 · 42.9%

Roughly 15–20% of the customers behind those 178 unconfirmed orders later came back and bought anyway, predominantly prepaid — on NAPPEE’s figures, somewhere around 27 to 36 people. The unconfirmed order is not simply a lost customer. It is an order that cost money and attention before anyone could tell whether it was real.

Two thirds of orders arriving COD is the number most brands would look at and conclude that COD is doing a great deal of work. That reading turns out to be premature.

The loss that never reaches a courier

Of the 415 COD orders, 178 were never confirmed, and so were never dispatched. That is 42.9% of COD orders and 29.2% of everything NAPPEE sold in the window.

A COD order was not necessarily a confirmed sale.

This distinction matters more than it first appears, because these 178 orders are invisible in the metric most brands actually watch. They never entered the courier network, so they produce no RTO record, no failed-delivery report and no return shipment. A brand tracking RTO alone would not see them at all.

That is not to say RTO was not also a problem. NAPPEE are clear that a meaningful percentage of the orders that were confirmed and dispatched did ultimately come back as RTO. They do not hold sufficiently reliable historical data to put a figure on it, so we are not going to estimate one, and neither should anyone reading this. The point is not that one leak is bigger than the other. It is that a brand watching only RTO is watching one of two.

What confirmation actually cost

Getting from a COD order to a confirmed one was not automatic, and it was not cheap. NAPPEE's customer-care team attempted confirmation twice a day for up to three working days per order, with WhatsApp reminders running alongside the calls.

The confirmation routine	As NAPPEE described it
Call attempts	Twice a day, up to three working days per order
Parallel channel	WhatsApp reminders alongside the calls
Team time	1–2 hours a day on calls, follow-ups and status reporting
Management time	Roughly a further 30 minutes a day of supervision
Side effect	Some customers used the confirmation call to negotiate a discount
Reachability	Their outbound number was flagged as spam by caller-ID services such as Truecaller

Two of those lines deserve pulling out.

The Truecaller problem is structural, not a NAPPEE problem. A number that dials many unfamiliar recipients in a short period can get flagged by caller-identification services, and once flagged, it is shown to customers as spam. The brand then has to reach people who have been told not to answer. Any Indian brand running outbound confirmation calls at volume is exposed to the same risk.

The discount negotiation is subtler and, we think, underrated. A confirmation call is an inbound-style conversation the brand initiated, which hands the customer an opening to renegotiate a price they had already accepted at checkout. The call is not merely a cost; it is a margin risk.

The arithmetic on staff time

The figures below are CentraTalk's arithmetic from NAPPEE's numbers, not NAPPEE's own reporting. We include them to give the effort a shape, not to put words in their mouth.

Derived figure	Calculation	Result
Daily effort, all in	1–2 hrs care + ~0.5 hr supervision	1.5–2.5 hrs/day
Across the 46-day window	1.5–2.5 hrs × 46 days	~69–115 hours
Ceiling on call attempts	415 COD orders × up to 6 attempts	up to ~2,490 dials

The 2,490 figure is a ceiling, not a count: most orders will have been resolved in fewer attempts, and NAPPEE did not report an actual dial total. But it frames the scale of the activity. Somewhere between two and three full working weeks of a person's time went into a 46-day experiment, and the largest single outcome of that work was 178 orders that never shipped.

The costs that are not staff time

Where a parcel did go out and come back, freight on the return ran roughly ₹100–250 depending on destination. That is before packaging, labels, receiving the parcel back and reconciling inventory — each small, each real, none of them on the courier invoice.

One cost NAPPEE raised is easy to miss and worth stating plainly: their platform's percentage-based fee could apply to a COD order even where nothing was dispatched and no sale was ever realised. A brand can therefore pay a transaction fee on an order that never became a transaction. Multiplied across 178 unconfirmed orders, that is a charge on pure non-revenue.

What they tried, and what happened

NAPPEE did not abandon COD without working the problem first.

Intervention	Outcome
OTP verification at checkout	Helped to a degree, but completing verification "was not a guarantee that the customer would ultimately accept the parcel"
Manual confirmation calls	Worked, but at 1–2 hours a day plus supervision, and the number was eventually flagged as spam
WhatsApp reminders	Run alongside the calls; did not remove the need for them
Disabling COD for high-return pincodes	Used as a blunt instrument to cut exposure
Converting a COD order to prepaid	No successful same-order conversion could be identified in the records they reviewed

That last row needs stating carefully. The standard industry answer to an unconfirmed COD order is to send a payment link and convert it on the spot. In the records NAPPEE reviewed, no successful same-order conversion of that kind could be identified.

That is not the same as saying conversion never happens. Roughly 15–20% of those customers did come back and buy later, mostly prepaid — and NAPPEE's records do not link a later prepaid order back to an earlier COD one, so there is no way to tell how many of those returns were, in effect, delayed conversions. What their data supports is narrower and still useful: the immediate save, on the same order, was not something they could evidence.

What happened when COD was switched off

Over the 46 days following the change, orders fell from about 13.2 a day to about 9.9 a day — a drop of roughly 25%.

NAPPEE were explicit that advertising, traffic and promotions could also have contributed to this decline, and that it should not be read as a clean measure of what COD was worth. Nothing was held constant. We are reproducing it because they were willing to share it, not because it settles the question. Anyone tempted to use 25% as a rule of thumb should not.

Customers still ask NAPPEE for COD today. Some pay prepaid once they are told it is unavailable. Some do not.

What would bring COD back

We asked NAPPEE what a solution would have to do for them to reconsider. Their answer reframes the problem, and it is the most useful thing in this study.

A large reduction in RTO, on its own, would not be enough. Any workable solution would also have to cut the manual effort: the confirmation calls, the repeated follow-ups, the tracking and reporting, the management supervision, the failed-delivery disputes and the return processing.

Our experience is that RTO alone doesn't capture the complete cost … the metric that matters isn't simply how many additional orders COD generates. It is how many of those orders ultimately become completed sales, and what it costs in money and staff time to get them there.

What we take from this

This section is CentraTalk's reading of NAPPEE's data. It is our interpretation, not their position, and they should not be held to it.

1. Measure the pre-dispatch loss separately from RTO. If 42.9% of COD orders never reach a courier, an RTO percentage is measuring one of two problems and is silent on the other. Any brand offering COD should be able to state, this month, how many COD orders were never confirmed — and most cannot, because no dashboard reports it as a category. Historical records did not allow NAPPEE to reconstruct a sufficiently reliable percentage for this research, which rather makes the point: the two numbers need to be tracked deliberately or neither is trustworthy.

2. The dominant cost is labour, not freight. Return freight at ₹100–250 is visible, invoiced and easy to attribute. One to two hours a day of a team's attention is none of those things, which is precisely why it is underweighted. When NAPPEE were asked what would bring COD back, they named the manual effort before anything else — which suggests the uninvoiced cost weighed at least as heavily as the invoiced one.

3. Anything that only reduces RTO is solving the wrong half. This is the implication of NAPPEE's last answer, and it sets a genuinely high bar. A tool that cuts RTO but leaves the confirmation workload intact does not change the decision they made. The work is the cost.

4. Outbound calling is getting harder, not easier. The spam-flagging problem can compound. The more diligently a brand chases confirmations, the greater the chance its number is flagged, and the harder the next confirmation becomes. Any approach built on dialling customers from a single outbound number is exposed to this, and it is unlikely to improve on its own.

Method and limitations

We are stating these plainly because the findings above are only as good as their basis, and the basis is narrow.

Limitation	What it means for these findings
One brand	This is a single case, not a sample. Nothing here generalises without corroboration.
One category	Baby diapers: repeat-purchase, mid-ticket. COD behaviour differs by category and price point.
46 days	A short window, and one that overlaps a festive and year-end period.
Self-reported	Figures were supplied by NAPPEE from their own records. We did not audit their systems.
Pre/post not controlled	The 25% decline has no control group. Advertising, traffic and promotions also varied.
Platform-specific	Fee behaviour on undispatched COD orders reflects the platform NAPPEE used, not all platforms.

What would improve this: the same questions put to five or six more brands, across different categories and price points, with the pre-dispatch confirmation rate reported as its own number. If you run a D2C brand in India and have that data, we would like to hear from you.

---

With thanks to NAPPEE for sharing their data and their time.