"Customer refused" is not a diagnosis

When a delivery fails, the courier logs a reason in the NDR (non-delivery report): customer refused, customer not contactable, address incomplete, customer asked for future delivery. Brands read these codes as facts. They are closer to claims.

Take "customer refused". That one code covers at least five different events:

1. 1The customer genuinely changed their mind. A real intent problem.
2. 2The customer was never told the parcel was coming and refused an unexpected COD demand. A communication problem.
3. 3The delivery executive called once, got no answer, and logged a refusal. An attempt-quality problem.
4. 4The delivery executive did not attempt at all and logged a refusal to close the day. This is common enough that courier platforms have a formal "fake attempt" dispute flow: Shiprocket's NDR process lets a seller contest an attempt by uploading a call recording or a chat screenshot from the buyer.
5. 5The customer wanted the product but not at that price or that day, and asked the delivery executive to take it back. A timing or price problem.

Four of those five are not "the customer did not want it". Each has a different fix, and none of them is fixed by tools that treat the NDR code as ground truth.

The reason codes persist because they cost nothing to log and they close the courier's ticket. Disputing them costs effort: recordings, screenshots, follow-ups. So the codes skew toward whatever ends the delivery executive's day.

GoKwik's analysis of its shopper network frames the same gap from the other side: it attributes 60 to 70% of RTOs to low buying intent (customer psychology) and 20 to 25% to logistics issues. Even if you accept that split, a quarter of RTOs being logistics-driven means a large block of orders that the customer wanted, failing inside the delivery process.

Two habits change the picture.

First, contact the customer yourself on every NDR, the same day, on WhatsApp or by call. Ask what happened. The answers will not match the codes often enough to surprise you, and each corrected NDR is an order rescued from the RTO pile. Courier platforms assume this: they allow only about three delivery attempts before an order is marked for return, so every wasted attempt is expensive.

Second, audit a sample. Take one week of NDRs, call twenty customers, and tally how many "refusals" were real. That percentage is more useful than any industry benchmark, because it is yours.

The NDR code is where the investigation starts. It is not where it ends.