Too Long? Read This in 10 Seconds
Executive bullet points for fast decision-making
- ✓ Three layers: what Meta charges, what your platform adds, and what the replies cost in staff time.
- ✓ Rates vary by message category and by country — a single blended number is always wrong.
- ✓ Judge platforms on the fee structure, not the headline price: seats, markups and expiring credits differ wildly.
- ✓ The metric that matters is cost per qualified conversation, not cost per message.
Every WhatsApp bill has three layers. Most business cases only count the first, which is why they are wrong.
Layer 1: what Meta charges#
Meta charges per message, and the rate depends on two things:
- Category — marketing, utility, authentication, and messages inside a customer service window are priced differently.
- Country — rate cards vary substantially between markets, sometimes by an order of magnitude.
Two rules for modelling this:
- Never use a blended global rate. Price your actual country mix and your actual category mix.
- Never hard-code rates into a spreadsheet you will not revisit. This is the fastest-moving part of the platform. Meta has changed both the charging model and the free allowances, most recently in the October 2026 update — the specifics of that change are covered in WhatsApp service message pricing and the 24-hour rule FAQ.
Always pull current rates from Meta's published rate card before signing anything.
Layer 2: what platforms add#
Providers sit between you and the API and charge for it. The models differ enough to change the answer entirely:
| Model | Looks like | Watch for |
|---|---|---|
| Per-message markup | A few paise or cents on top of each message | Applies to every message, including ones that fail |
| Per-seat | Monthly fee per agent | Cost scales with team size, not volume |
| Tiered plans | Message bundles with a monthly floor | Unused bundle expiring monthly |
| Pay-as-you-go wallet | You top up, messages debit | Check whether credits expire |
| Hybrid | Platform fee plus markup | Read both carefully |
Questions worth asking before you commit:
- Are failed messages charged?
- Do credits expire, and after how long?
- Is there a minimum commitment or annual lock-in?
- Is template management, campaign sending or the API included, or extra?
- What happens if you exceed a plan mid-month — throttling, overage, or a hard stop?
- Can you export your contacts and conversation history if you leave?
The last one is the important one and the one almost nobody asks.
Layer 3: the replies#
The layer no spreadsheet includes. A campaign that generates a thousand replies needs people to answer them, and unanswered replies convert worse than sending nothing.
If a campaign produces 1,000 replies and an agent handles 40 conversations an hour, that is 25 agent-hours. Cost that at your loaded hourly rate and add it to the campaign. It is frequently larger than the message spend.
A worked model#
Do this arithmetic with your own numbers rather than trusting anyone's benchmark:
Campaign inputs
Recipients R
Marketing rate (your country) M per message
Platform markup P per message
Reply rate r
Qualified rate (of replies) q
Close rate (of qualified) c
Average order value V
Agent minutes per conversation t
Loaded agent cost per hour H
Costs
Message cost = R × (M + P)
Staffing cost = R × r × t / 60 × H
Total cost = message + staffing
Returns
Qualified convos = R × r × q
Orders = qualified × c
Revenue = orders × V
Report
Cost per qualified conversation = total cost / qualified convos
Contribution = revenue − total cost
Two habits that keep this honest:
- Divide by qualified conversations, not by replies. "Not interested" is a reply.
- Compare against the counterfactual. Some of those orders would have happened anyway. Holding out a control segment is the only way to know how many.
Cost controls that survive a growth month#
- Frequency caps per contact, enforced in the sending path.
- Segment floors — never send to a segment you cannot justify individually.
- Category discipline. Sending a utility-eligible message as marketing costs more and converts no better.
- List hygiene. Every dead number costs full price to message.
- Budget alerts at 50%, 80% and 100% of the monthly plan.
- Reply capacity check before scheduling — if nobody can answer on Saturday, do not send on Saturday.
Against SMS and email#
- Per message: SMS is often comparable or cheaper; email is far cheaper.
- Per delivered-and-read: WhatsApp leads comfortably.
- Per reply: WhatsApp is not close to being matched — conversations happen, which is the entire point.
- Per closed order: depends on whether you staff the replies. Unstaffed, WhatsApp's advantage disappears.
The honest summary: WhatsApp costs more per message and usually less per outcome, provided someone answers.
See AR-Inbox pay-as-you-send pricing — per-message billing with a wallet, no seat fees, and per-campaign cost reporting.
Related: the October 2026 pricing change · metrics worth reporting.
Frequently asked questions
How much does the WhatsApp Business API cost? +
Meta charges per message, with rates that vary by message category and destination country, and your platform provider adds its own fee. Always price your own country and category mix rather than a blended average.
Is the WhatsApp Business API free? +
Access is not sold by Meta as a subscription, but messages are charged. Customer-initiated conversations are cheaper to serve than business-initiated ones, and free allowances have narrowed over time.
How is a WhatsApp conversation charged? +
Charging has moved toward a per-message model by category. Check the current Meta rate card for your market before modelling — this is the fastest-changing part of the platform.
Is WhatsApp cheaper than SMS? +
Per message, often not. Per outcome, usually yes, because reply rates and conversion rates are higher. Compare cost per qualified conversation, not cost per send.
WhatsApp policies, limits and pricing change often. This article was last reviewed on Aug 21, 2026 and is scheduled for its next review on Sep 30, 2026.