The WhatsApp 24-hour service window, explained properly

If one rule explains most of WhatsApp Business API behaviour that surprises people, it is this one. Your reply failing to send, your message costing money when you expected it to be free, your team being told to use a template for what felt like a normal conversation — all the same rule.

What it is

When a customer sends your business a message, a 24-hour customer service window opens on that conversation. While it is open, you can reply with anything: free text, images, PDFs, voice notes, buttons, a product list. No template, no approval, no charge.

When it closes, you can only send an approved template, and templates are billed.

That is the whole rule. Everything else is a consequence of it.

How it opens, resets and closes

• Opens when the customer sends you a message.
• Resets to a fresh 24 hours every time the customer sends another message. It is not a rolling budget you spend down — one new inbound message and the clock starts over.
• Closes 24 hours after their last message, not after your first reply.
• Your messages do not extend it. You could reply twenty times; the window still expires 24 hours after they last wrote.

A conversation that runs over a week, with the customer replying every day, has an open window the whole time. A customer who asks a question on Friday evening and hears nothing until Monday is behind a closed window, and now the cheapest possible follow-up costs you a template.

What it costs

Free-form replies inside the window are service messages, and service messages are free.

There is a second, less-known freebie: a utility template sent while the window is open is also not billed. So an order-update template to someone who messaged you an hour ago costs nothing; the same template two days later does.

Marketing templates are billed whether the window is open or not.

The practical version:

You send	Window open	Window closed
Free-form reply	Allowed, free	Not allowed
Utility template	Allowed, free	Allowed, billed
Marketing template	Allowed, billed	Allowed, billed
Authentication template	Allowed, billed	Allowed, billed

Which makes speed of reply a cost control, not just a service metric. A team that answers within the hour pays for almost nothing. A team that answers in two days pays for everything.

What this does to how you work

Surface the timer where people reply. The single most common operational failure is someone writing a careful free-form reply to a conversation whose window closed overnight, and only discovering it when the send fails. Watx shows the window counting down in the inbox on every conversation for exactly this reason.

Have a utility template ready for the common follow-ups. Order shipped, appointment tomorrow, document received. These are the messages that need to go out on a schedule rather than in a conversation, and they need approval in advance.

Do not use marketing templates to reopen conversations casually. It works — they are allowed at any time — but it is the expensive way, and it spends the goodwill that keeps your quality rating healthy.

Route round the clock if your customers message round the clock. This is where an AI agent earns its place: not to replace a person, but to keep the window from closing on a question nobody was awake for. A reply at 2am is free, and it resets nothing — but it means the customer is still in conversation with you at 9am rather than waiting on a template.

One thing people get wrong

The window is per customer, not per business, per number or per agent. Every contact has their own clock. There is no global state to manage — only a per conversation one, which is why it belongs in the inbox UI rather than in a report.

---

Related: what Meta charges per message · why templates get rejected