What the WhatsApp Business API
actually costs to run.

Almost every WhatsApp automation quote we are asked to review prices the build carefully and the running cost barely at all. That is the wrong way round. The build is paid once. The messaging bill arrives every month for as long as the system is useful, and it is the number that decides whether anyone still uses the thing in year two.

Meta bills per message now, not per conversation

On 1 July 2025 the model changed. The old system charged for a 24-hour conversation window; the new one charges for each delivered template message, priced by its category, the recipient's country, and your monthly volume tier. If a quote you are reading still talks about conversation pricing, it was written against a rate card that no longer exists.

The practical effect is that message design now moves money. Three separate templates where one would do used to cost nothing extra inside an open window. Now it costs three times as much, every time, forever.

The category is the price

Every template is classified, and the classification — not the length or the content — sets the rate.

Two consequences worth designing around. First, service messages are free: if a customer messages you first, everything you send back for the next 24 hours costs nothing. A support flow built to reply inside that window rather than to push templates at people can run at close to zero messaging cost. Second, volume discounts apply to utility and authentication but not to marketing. Meta has deliberately made bulk promotional sending the one thing that never gets cheaper.

The cheapest WhatsApp system is usually the one that answers people rather than the one that broadcasts at them. That is a design decision, and it is made before any code is written.

The line items quotes leave out

1. The BSP markup. Meta's fee is identical whichever Business Solution Provider you go through. The provider's markup on top is where they differ, and it is commonly in the range of USD 0.003 to 0.010 per message — which on utility traffic can be several times Meta's own charge. Ask for it as a separate line, not folded into a bundled per-message price.
2. Template rejection cycles. Templates are reviewed before they can send. Rejections are routine early on, and each one is a round trip of a few hours to a day. Budget the calendar time, not just the effort.
3. Quality rating management. Numbers carry a quality rating driven by user blocks and reports. A rating that falls restricts your sending limits. Somebody has to watch it, and if nobody does you find out when a campaign silently throttles.
4. Opt-in records. You need a defensible record of who agreed to be messaged and when. That is a small piece of engineering that is invisible until a complaint arrives, at which point it is the only thing that matters.
5. Failure handling. Messages fail. Numbers change, users block, rates limit. A system without a retry and dead-letter path quietly loses messages and nobody notices for weeks.

The twenty-times trap

Authentication messages to Indian numbers are among the cheapest traffic on the platform. Authentication messages to international numbers from an Indian sender are not — the reported gap is roughly twenty-two times. A login flow that works fine for a domestic user base becomes a different business case the moment you have overseas customers, and the change is invisible in testing because the code is identical.

If any part of your user base is outside India, model the OTP cost per market separately before you commit to WhatsApp as the delivery channel. Sometimes the honest answer is SMS for one market and WhatsApp for another.

How to estimate your own bill

Four numbers, and you can do it on paper. Monthly volume per category; the rate for each category in your recipients' markets; your BSP's markup; and the proportion of your traffic that can be moved inside a free service window by designing for replies rather than pushes. That last one is usually the largest single lever, and it costs nothing but thought.

Then add fifteen to twenty per cent for the messages you have not thought of yet. Every system we have built sends more than the first estimate, because once it works people find uses for it.

What we do about it

We quote the monthly running cost alongside the build, before you commit, with the categories broken out and the BSP markup shown separately. If the running cost makes the project a bad idea, that is worth discovering in a spreadsheet rather than in month four.