The free reply was the single most important fact about WhatsApp economics for order taking. Meta has published the date it ends.
If you have read anything about WhatsApp economics for B2B ordering — including earlier posts on this site — you have read some version of this: retailers message you first, that opens a 24-hour window, and everything you send inside it is free. Build flows that earn replies, because replies are free.
That advice was correct. Meta has now published the date it stops being correct.
According to Meta's pricing documentation for non-template messages, from 1 October 2026 it will charge on a per-message basis for service messages, consistent with how it charges for template messages. The same date applies to utility messages sent in response to users inside an open 24-hour customer service window. Both categories are free until then.
There is a second, separate change dated 1 August 2026: Meta Business Agent messages are charged on a per-token basis, at a published rate of $2.00 USD per 1M tokens. That one is about AI-generated replies specifically and is covered in what an AI reply actually costs per conversation.
Worth being precise, because the change is narrower than the panic around it.
The window still exists as a permission rule. Inside 24 hours of a retailer's message you may send free-form content — text, images, PDFs. Outside it you must use an approved template. That distinction is unchanged.
Template pricing is unchanged. Marketing and utility templates you initiate were always billed per message and continue to be.
Categories still matter. Marketing remains the expensive category and utility the cheap one. The mechanics of category assignment and why a mis-categorised template costs you are in WhatsApp template categories and what they cost.
What changes is one thing only: being inside the window stops making a message free. From October, the window tells you what you are allowed to send, not what you pay for it.
A broadcast-style business sends one template and is done. Its cost barely moves.
A conversational ordering business is the opposite. Taking an order over chat is inherently multi-turn: the retailer asks, the bot answers, the retailer adds an item, the bot confirms the cart, the retailer approves, the bot sends the invoice. Six messages, all inside the window, all free today.
From October, that is six billed messages for one order. A flow that does the same job in two is not marginally better, it is a third of the cost — and the difference is entirely in design, not in what the retailer receives.
This inverts an assumption worth naming. The old advice was to earn replies, because a reply opened a free window. That advice is now half right: a reply still opens the window you need for permission, but the conversation inside it is no longer free, so length is no longer costless.
Numbers to re-run with your own rate card, not a benchmark. The rate below is a placeholder — substitute the current published rate for your market.
Take 400 retailers ordering twice a month, so 800 orders. Call the per-message rate R.
Under a six-turn flow, that is 800 × 6 = 4,800 billed messages a month, or 4,800R.
Trim the same flow to three turns — a catalogue-aware first reply that confirms the whole cart at once, then a confirmation, then the invoice — and it is 2,400 messages, or 2,400R. Same orders, same retailers, half the bill.
| Flow design | Messages per order | Monthly messages | Relative cost |
|---|---|---|---|
| Conversational, unoptimised | 6 | 4,800 | 100% |
| Consolidated confirmations | 3 | 2,400 | 50% |
| Single-turn reorder for regulars | 1 | 800 | 17% |
That third row is the interesting one. A retailer reordering their usual list does not need a conversation at all — they need one message they can approve. The design work is knowing which retailers are in that pattern, which is the reorder cover question applied to messaging.
Count your turns. Take twenty recent order conversations and count messages per completed order. Most teams have never measured this because it never cost anything. That number times your rate card is your exposure.
Collapse acknowledgements. The cheapest messages to remove are the ones carrying no information — a bare "got it", a separate "adding to cart" for each line. Fold them into the next substantive reply.
Confirm carts once, not per line. A retailer sending five items should get one cart summary, not five confirmations. This is usually the single biggest reduction available and it also reads better.
Move the regulars to one-tap reorder. For a retailer whose order barely varies, send one message with the usual list and a confirm action. One billed message replaces six.
Re-check where you send PDFs. Invoices and statements are messages too. Batching a weekly statement instead of sending per-invoice copies is worth checking against how retailers actually use them.
Leave the permission design alone. None of this changes opt-in obligations, which are a separate constraint and arguably a harder one — see WhatsApp opt-in and what counts as consent.
The tempting reaction is to push retailers off WhatsApp and onto an app or portal to avoid per-message costs. For B2B distribution in India that trade is almost always bad, and the reason is in how Indian SMBs actually adopt technology: the entire reason WhatsApp ordering works is that the retailer already has it open and needs no training. An app that saves a few rupees per order and loses a fifth of your retailers to friction is not a saving.
The correct response is to make the conversation shorter, not to move it somewhere worse.
Two habits worth adopting, given how much secondary commentary this change has attracted.
Rates and dates come from Meta's own pricing documentation, which is versioned and dated. Reseller and aggregator blogs quote rates that were current when written and are frequently stale. Meta's stated practice for these updates is to give more than a month's notice, so the authoritative page is the one to check before you budget.
And confirm what your own provider adds. Meta's rate is the floor; a BSP or aggregator may add a markup per message on top of it. The number that matters for your planning is the one on your invoice, not the one in the documentation.
The questions that come up once a team realises the free reply is ending.
Meta documentation states that from that date it will charge on a per-message basis for service messages, consistent with how it charges for template messages, and will also charge per message for utility messages sent in response to users inside an open 24-hour customer service window. Both of those are free today, which is why this is a change in kind rather than a rate adjustment.
Yes. The window still governs what you are permitted to send — free-form replies inside it, approved templates outside it. What changes is that being inside the window no longer means the message is free. Treat the window as a permission rule from October onward, not as a pricing rule.
Meta publishes rates per market and by category, and the documentation for this change directs you to the rate card rather than naming a single global number. Because rates are revised periodically, the only reliable figure is the current rate card for your market at the time you plan. Work out your own exposure by multiplying your monthly reply volume by that rate rather than by any number quoted in a blog post, including this one.
Cut the number of messages it takes to complete an order. Under the old model a six-turn conversation and a one-turn conversation both cost nothing inside the window, so nobody optimised for turn count. Once every reply is billed, a flow that confirms an order in two messages costs a third of one that takes six, for exactly the same outcome.
The FlowKartAI team builds WhatsApp-native ordering for Indian B2B distributors and the kirana stores they serve. We write about distribution economics, GST compliance, and the practical side of putting AI in front of retailers who have never opened an app.
FlowKartAI parses natural language WhatsApp messages into ERP-ready orders in seconds.
Try Free Demo