How to Confirm COD Orders Before Dispatch in Bangladesh
Roughly three quarters of online orders in Bangladesh are paid in cash at the door. That single fact changes what an order is. In a prepaid market an order is money that has already moved. Here it is a promise, and the seller carries the whole cost of a broken one: the forward leg, the return leg and the handling.
Confirmation is the step that turns a promise back into something closer to a commitment. It is also the step most often done badly, done too late, or skipped on exactly the orders that needed it. This is the operational version: when to confirm, what to say, which channel to use, and what to do when nobody picks up.
Should you call every cash-on-delivery order before dispatch?
No. Confirming every order costs more than it saves, and it delays the orders that were never in doubt.
The orders worth confirming are the ones carrying a signal: a first-time buyer, a high value against your average, an address that resolves to a landmark rather than a house number, a phone number that has refused a parcel before, or an order placed at an hour when impulse runs ahead of intent.
A returning customer with three accepted deliveries and a saved address does not need a call. Calling them adds a day to a delivery they were always going to accept, and a day of delay is itself a reason people cancel.
The decision is therefore not whether to confirm but which orders to confirm, and that is a scoring question rather than a policy question.
When is the best time to confirm a COD order?
Inside the window where the order can still be changed, and before the courier consignment is created.
This is the detail that catches most sellers. Once a consignment exists, Pathao and Steadfast expose no way to cancel it and no way to change the cash-on-delivery amount. Whatever you learn after that point, you cannot act on it. The parcel goes.
So the confirmation has to happen in the gap between the order landing and the booking being made, which is why holding orders to a daily cutoff rather than dispatching them one by one is a deliberate choice rather than a lazy one. It creates the only window in which a correction is still possible.
In practice that means confirming within a few hours of the order, and batching dispatch to a cutoff late enough to have heard back.
What should an order confirmation call actually say?
Four things, in this order: who is calling, what was ordered, what it will cost at the door, and when it will arrive. Then one closed question.
The common mistake is opening with a question that invites reconsideration. Do you still want this is an invitation to say no. Confirming your order of two items for 1,450 taka, arriving Thursday, is a statement that invites a correction instead.
Read the total including delivery. A customer who expected 1,200 and is asked for 1,450 at the door refuses the parcel, and that refusal was created by your own silence rather than by the customer.
Confirm the address out loud. In a market where a great many addresses are given as landmarks rather than house numbers, reading it back is the cheapest correction you will ever make.
Keep it under sixty seconds. This is a verification, not a sales call, and treating it as a sales call is how confirmation calls start getting ignored.
What do you do when the customer does not answer?
Do not treat silence as a refusal, and do not treat it as a confirmation either. Escalate down a channel ladder and give it a defined window.
A missed call in Bangladesh means very little on its own. People are at work, on another line, or screening an unknown number. What it does mean is that you have not yet got the signal you were looking for.
A workable ladder is one call, then a written message on the channel the customer originally used, then one more call at a different time of day. If the order still has not been confirmed by your cutoff, the decision is not to cancel it. The decision is to route it differently, because an unconfirmed order is not worthless, it is riskier.
That is the part most guidance gets wrong. Unconfirmed does not mean cancel. It means send it with the courier most likely to complete a difficult delivery rather than the cheapest one, and accept a slightly higher delivery charge in exchange for a materially better chance of collecting.
Should you confirm by call, WhatsApp or Messenger?
Confirm on the channel the customer already used, and fall back to a call.
A buyer who placed the order in your Messenger inbox is reachable in your Messenger inbox. Moving them to a phone call adds friction and an unknown number. The reverse is also true: someone who phoned the order in is not going to read a template message.
There is a cost dimension as well. Since 1 July 2025 Meta prices WhatsApp per message rather than per conversation, and utility messages sent inside an open customer service window are free. A confirmation sent while the customer is still in conversation with you costs nothing; the same message sent three days later as a template does not.
Voice earns its place where the order is genuinely uncertain, because a call gets an answer in one exchange and a message may not get one at all.
Can order confirmation be automated in Bangla?
Yes, and the language is the harder half of the problem rather than the automation.
A confirmation is a narrow, scripted, high-volume interaction, which is the kind of work automation handles well. What makes it difficult in this market is not the script, it is that a Bangladeshi customer answering the phone will reply in Bangla, in English, or in both inside a single sentence, and a system that has to pick a language before it starts listening will mishear the answer.
The practical test for any automated confirmation, whether you build it or buy it, is a code-switched sentence. Ask the vendor to demonstrate one live rather than describing it, because a translation layer cannot fake a sentence that changes language halfway through.
Omnistra runs this step as part of Order OS: the risk decision, the confirmation itself in Bangla or English, and the courier booking sit in one flow rather than in three tools with a person copying between them.
How do you know whether confirmation is working?
Measure the delivery success rate of confirmed orders against unconfirmed ones over the same period and the same areas.
One number is not enough on its own. Confirmation that reaches nobody looks identical in a report to confirmation that reaches everybody, so track contact rate separately from delivery rate. If your contact rate is low, the problem is the channel or the timing, not the script.
Watch cancellation-at-confirmation too, and do not treat it as a failure. An order cancelled on the phone costs you a minute. The same order cancelled at the door costs you both legs of the journey plus handling, and that is the trade the whole exercise exists to make.
