When not to migrate to the official WhatsApp API
By José Pedro · Co-founder of CRPRO
The official API isn't the right call for everyone. See the four cases where it doesn't pay off, what changed on October 1, 2026 for support-heavy operations, the real difference in risk between QR code and the official API, and the five numbers to gather before deciding.
The official WhatsApp API — Meta's WhatsApp Business Platform — doesn't pay off in four situations: very high inbound support volume, a contact base with no opt-in, a one-person operation with low volume, and use that depends on groups, status, or the phone app. Outside those four cases, migrating usually pays off. The list of exceptions grew on October 1, 2026, the date Meta started charging per message for what used to be free inside the 24-hour window. CRPRO sells both models, the official API and QR code, so there's no reason here to push you toward either.
In which cases isn't it worth migrating to the official API?
| Case | Why |
|---|---|
| Very high inbound support volume, few campaigns | Since October 2026, every reply inside the window is a charged message, with no volume discount |
| You have nobody's opt-in | With no logged permission, there's no template to send |
| A one-person operation, low volume | The API's gain is process. With no team and no campaigns, there's nothing to organize |
| Your use depends on groups, status, or the phone app | A migrated number leaves the app. Groups and status don't exist on the API |
If none of these is you, migrate. If one of them is, keep reading before deciding. The four weigh differently: the first depends on the service rate in Meta's price table, and the other three are already true today, with or without October's change. The fourth is the only one that's practically irreversible.
What changed on WhatsApp on October 1, 2026?
What's already changed and a lot of people haven't updated on: since July 1, 2025, Meta charges per delivered template message, no longer per conversation. Any table with a fixed per-conversation price describes a dead model.
| Message type | Through September 30, 2026 | Since October 1, 2026 |
|---|---|---|
| Service message (your reply to someone who reached out) | free | charged per message, no volume discount |
| Utility template inside the open window | free | charged per message, with a volume discount |
| Any non-template message inside the window | free | charged per message |
| Everything inside the Click-to-WhatsApp 72-hour window | free | free, no change |
| Marketing template | charged | charged, no change |
Two rows in that table decide everything. The first: service gets no volume discount, so support cost grows in a straight line — double the volume, double the bill. The fourth: whoever brings conversations in through Click-to-WhatsApp ads keeps the 72-hour breathing room. If you want the full breakdown of who feels the hit most, it's in October 2026's charge and who feels it.
How much does a message on the official API cost today?
The service-message rate is in Meta's current price table, and anyone showing a locked-in figure without citing it is guessing. What you can use as a reference is the template price in effect in Brazil as of August 2026, inside the current 24-hour window.
| Template category | Rate in Brazil (in effect as of August 2026) | Approximate in reais |
|---|---|---|
| Marketing | US$0.0625 per delivered message | around R$0.32 |
| Utility and authentication | around US$0.0068 per delivered message | around R$0.035 |
The utility and authentication row is much less solid than marketing's: use it as an order of magnitude, don't build an argument on top of it. BRL billing has been available since July 1, 2026 and becomes mandatory by June 30, 2027, which changes currency exposure, not the dollar price.
Case 1: how much does it cost to handle support per message?
Count what matters: it isn't a conversation, it's a message. The customer asks about price, timing, and the Pix code — three service messages in one conversation, and business support rarely fits in fewer than six or eight replies.
Multiply: 3,000 conversations a month at 8 replies each is 24,000 service messages. With no rate quoted here, all you can show is the slope of the curve per cent:
| Service messages/month | Cost per R$0.01 of rate | If it's R$0.02 | If it's R$0.05 |
|---|---|---|---|
| 2,000 | R$20 | R$40 | R$100 |
| 24,000 | R$240 | R$480 | R$1,200 |
| 60,000 | R$600 | R$1,200 | R$3,000 |
The last two columns are invented hypotheses to show the shape of the curve. They aren't Meta's rate — the real service rate is in Meta's current price table. Swap it in.
The structural point is the middle column: with no volume discount on service, this line never doubles downward. A low-ticket, long-conversation operation — a school, a real estate agency, an internet provider — sees costs grow along with the customer base, on the metric that generates no new revenue. On QR code, that same volume costs zero per message: just the platform's monthly fee.
Where the official API's math still adds up easily
| Profile | Why October 2026 weighs little |
|---|---|
| Whoever runs campaigns | Marketing has always been paid; nothing changes on that line |
| Whoever advertises via Click-to-WhatsApp | The free 72-hour window is still standing — it's the biggest exemption left |
| Whoever sends transactional notifications at volume | Utility and authentication keep their volume discount |
| Whoever has a team | With three or more people on the same number, multi-agent support usually already pays for the difference |
The pattern across all four profiles is the same: the official API's value isn't in the individual message, it's in sending at scale with permission, or in organizing people around a single number. Whoever has either of these two problems keeps their math working after October. Whoever has neither is the one who needs to redo the arithmetic with the rate in Meta's price table.
Case 2: can you migrate with no opt-in on your base?
Not the way most people would like. The official policy requires opt-in: you can only contact someone who gave you the number and gave permission. A purchased spreadsheet doesn't turn into permission just because you checked a box in your own system.
With no opt-in, the only way to open a conversation is a template — and a template sent to someone who didn't ask for it turns into a block and a report. That's exactly the behavior feeding the number's quality score, which Meta calculates on a weighted basis over the last 7 days. Migrating doesn't solve the problem of a list with unclear origins: it moves the problem into an environment where it's measured, and where the measurement has consequences on your sending limit. Build the permission first, migrate afterward.
Case 3: is it worth migrating as a one-person operation?
The API doesn't improve the message, it improves the process: queueing, distribution, history, an approved template, reporting. Whoever handles 40 conversations a month alone, from their phone, has no process to organize — the monthly fee and the per-message cost come in with no upside at all.
The practical cutoff is the team, not revenue. With three or more people fighting over the same number, what you gain in organization usually already covers the cost today. Alone and with low volume, it almost never does. And the gap widened in October 2026: every reply inside the window now has a price, while the process gain stays zero for whoever has no process.
Case 4: does the number keep working in the app after migrating?
No, and this is the cost that catches the most people off guard. When a number migrates to the official API, it stops working inside the WhatsApp app. Gone along with it:
- groups — the API doesn't send or receive in groups;
- status updates — they don't exist on the API;
- the prior history — it doesn't come along.
If your operation lives off a customer group, or if whoever handles support needs the app in hand to work, this isn't an implementation detail. It's a reason not to migrate that number. Worth noting the decision is per number, not for the whole company: whoever depends on a group on one number and runs campaigns on another has an out, and it's below, in the middle path.
What QR code costs in exchange for zero fees
QR code has no per-message cost, works with groups, and requires no template approval. The price is risk: on the regular app you can be banned with no warning at all, no quality score, and no formal appeal. You find out when it stops opening.
And there's no immunity on the other side either. The official API doesn't eliminate ban risk — Meta keeps the waba_ban_state field in the account webhook, and error 368 still exists. A number on the official API can be banned. Anyone selling you the opposite is lying.
What's the real risk difference between QR code and the official API?
The difference isn't immunity, it's visibility and due process.
| QR code | Official API | |
|---|---|---|
| Per-number quality score | doesn't exist | green/yellow/red, calculated over the last 7 days, visible in WhatsApp Manager and via the API |
| Warning before the penalty | none | a quality alert by email before restriction |
| Escalation | straight to blocking | sending limit restricted before blocking |
| Appeal | informal, no channel | a formal contest channel |
| Sending outside the window | nothing stops it — this is where numbers get burned | impossible: returns error 131047, only a template reopens it |
That last row dismantles a myth: on the API, sending outside the window isn't a "cause of a ban," it's impossible — window risk is a QR code problem. If you want to understand how the score works before migrating, see WhatsApp quality rating. If you've already had a number burned, it's worth reading what to do when WhatsApp bans you and what WhatsApp considers spam. And the obvious point: GB WhatsApp and WhatsApp Plus violate the Terms and aren't the third way they seem to be.
The five numbers to gather before deciding
With Meta's price table in hand, you'll need these five numbers to decide in an afternoon instead of a quarter:
- Inbound conversations per month.
- The average number of messages your team sends per conversation — this is the multiplier, and almost nobody knows their own.
- Utility templates per month (confirmation, delivery, billing, reminder).
- Marketing templates per month.
- What it costs to lose the number: the contact base, the history, the number printed on the storefront and in ads.
The fifth item breaks the tie. Losing with no warning a number that's been on the storefront for years weighs more than any per-message cent difference — and it's the only variable on the list that doesn't reset the following month.
The middle path: both models at once
The operation doesn't need to sit entirely on one side. A common setup: the official API for the number that sends campaigns and notifications — where bulk sending requires an approved template anyway — and very high-volume support on QR code. Opt-in matters on both: QR code isn't permission to bulk-send.
The other half of the math is the platform's monthly fee, on plans, broken down in how much the WhatsApp Business API costs. One hidden cost to close the loop: the Meta Business Agent has been billed per token since August 1, 2026, at US$2.00 per million tokens. Meta's documentation says a message typically consumes 20,000 to 25,000 tokens — which works out to 4 to 5 US cents per interaction. The full breakdown is in how much the Meta Business Agent costs.
On CRPRO you can set up both models and switch later, without starting from scratch.
Frequently Asked Questions
Does the official API protect my number from a ban?
No. It gives you visibility and due process — a per-number quality score calculated over the last 7 days, an email alert before the penalty, and a formal contest channel. A ban stays possible: Meta keeps deactivation states in the account webhook and error 368. The API eliminates the surprise, not the risk.
How much does the service message cost?
It's in Meta's current price table, and any figure circulating without citing it is a guess. What's confirmed here is only the structure — per-message billing, with no volume discount for service.
Is it worth migrating just to handle support better, with no campaigns?
Depends on team size. With three or more agents on the same number, the organizational gain usually covers the cost. Alone and with low volume, almost never — even more so since October 2026, when every reply inside the window is charged.
Can I go back after migrating?
You can, but it hurts. The number leaves the app once it joins the API, prior history doesn't come along, and undoing it involves removing the number from the business account. Treat it as a mid-term decision.
Do groups work on the official API?
No. The API doesn't send or receive group messages, and status updates don't exist on it either. If your operation depends on a group, that's a strong reason to keep that number on QR code.
What happens if I try to reply after 24 hours on the API?
The send fails with error 131047. Only an approved template reopens the conversation. That's why "sending at the wrong hour" is a QR code risk, not an API one.