Meta has published the October 1 rates, so the remaining input is your own message history: how many messages the customer service window currently makes free, and where their recipients are.
The method
- Take the last 30 days of outbound messages. A full month smooths out weekday and campaign effects.
- Keep only the ones the customer service window made free. Include free-form service replies and utility templates delivered inside the window. Exclude messages made free by an ads entry point.
- Bucket them by recipient country. Rates differ per market, so a single total will mislead you.
- Multiply each bucket by that market's rate.
The first three steps are the work, and the next section covers where to count from. The last one is a multiplication, and the calculator uses Meta's published October 1 utility rates:
Prefilled with 12,000 messages a month, split 70/30 between Mexico and Brazil. Replace the markets and volumes with your own. Service messages and utility templates inside the customer service window use the published October 1 utility rate. This estimates Meta list-price spend before volume discounts; it is not an invoice quote.
Counting from the right source
Count from status webhooks, not from send calls. A send log tells you how many messages went out; only the status data tells you which were free, and the free ones are what you're counting. Chapter 0.5 has the mechanics.
If you've kept the raw pricing payloads, start with outbound messages Meta marked free because of the customer service window, then retain both the service and utility categories. If you haven't kept them, combine free-form replies with utility templates delivered while a window was open.
Replies inside an ad's 72-hour entry-point window come back as free_entry_point, not free_customer_service, so this count leaves them out. That's correct: they stay free after October, and for a business where much of the volume starts from ads, the estimate comes out meaningfully lower because of it.
Two details matter however you build it. Count distinct message IDs, not status rows, because one message produces several statuses and Meta redelivers webhooks; each message should count once. And bucket by the first two letters of recipient_user_id, which are the recipient's country. Every status carries them, with or without a phone number, so there's no dialling-code table to build.
If your number runs on Kapso
Open your project's Usage screen and switch to the WhatsApp tab. The Meta service fees card projects service message volume separately for each WhatsApp number and applies up to 1,000 free messages per number per month. Unused allowances are not transferred between numbers; each number’s allowance is split proportionally across its markets. Select Calculation for the market breakdown and calculation assumptions. Add utility templates sent inside the window to that number; the calculator above prices the combined volume at the published October rates.
What it adds per conversation
Before multiplying anything by 100,000, look at what October adds to a single conversation.
Take a support conversation where you send six replies. From October 1, each of those bills at the recipient market's utility rate, so at the published rates the whole conversation costs:
- about $0.005 more in Colombia (rate $0.0008),
- about $0.05 more in Mexico (rate $0.0085),
- about $0.16 more in Argentina (rate $0.0260).
Now put that next to what you already spend producing the replies. If an AI agent handles the conversation and the model usage costs you $0.50, Meta's fee adds about 10% in Mexico and about 1% in Colombia. If a human agent handles it, their time costs dollars per conversation, and the fee barely moves the total.
So if answering already costs you something, no single conversation gets meaningfully more expensive. What makes October a budget item is volume. 100,000 of those conversations a month in Mexico is about $5,000 a month that didn't exist before. The calculator above works that out for your own volumes.