WhatsApp does not grade your copy. It grades how people react to your messages: blocks, reports, and reads, rolled into a 7-day quality rating that decides how many conversations your number may start each day.
The protection is behavioural, not clever wording. Never message a bought list, put an unsubscribe on every marketing send, and build months of wanted utility traffic (confirmations, receipts, reminders) before you broadcast anything promotional.
A number with a deep history of clean utility sends can survive one bad campaign. A cold number cannot. This guide covers the full playbook we run on client lines.
Here is something Meta never puts in a big red banner: the rules that actually get WhatsApp numbers throttled and banned are scattered across policy pages, and most business owners meet them for the first time after their number is already limited.
We manage WhatsApp lines for clients every day and have sent 15M+ broadcasts across client campaigns. What follows is not the official documentation rewritten. It is the operating playbook that has kept those lines healthy, with the reasoning spelled out so you can apply it yourself.
How Meta actually judges your number
Meta grades your WhatsApp number on reactions, not content. What it measures is behaviour around your outgoing messages: how many you send, how many get delivered and read, and above all how many people tap Block or Report after receiving one.
Those reactions feed your quality rating: a traffic light (green, yellow, red) reflecting roughly the last 7 days of feedback. The rating sits behind a second number that matters even more, your messaging tier: a daily allowance of people your number may start conversations with, which grows in steps as you prove quality at volume (Meta's messaging limits).
The loop is simple and unforgiving. Good reactions raise the allowance. Blocks and reports drop the rating, and a low rating steps the allowance back down and freezes upgrades. Since late 2025 that allowance is shared across your whole business portfolio, so one reckless number drains the pool for every number you run.
This is the single most useful mental model: every outgoing message is a small vote on your number's future, cast by the person receiving it.
The four steps between a warning and a ban
Meta rarely kills a number in one move. Enforcement escalates through four distinct steps, and each one is a chance to stop and fix the cause.
The phrase our team trains on: a ban is a pattern, not one slip. It is volume that outran the consent and quality behind it, sustained past the warnings. Which means every rule below is really about never letting the pattern start.
Rule 1: never buy a list
A bought list is the fastest known route to a dead number. Every contact on it is a stranger who never agreed to hear from you, which makes each send a coin flip between being ignored and being reported.
Run the maths from the receiver's side. A thousand people get a promotion from a business they have never spoken to. Some fraction will block on sight, some will report "did not sign up", and Meta reads both signals within days.
A cold number has no history to absorb that damage, so the rating flips fast and the allowance steps down before the campaign even finishes. The list seller keeps the money; you lose the line. There is no volume of sales from a bought list that outruns this arithmetic.
Rule 2: always give people a way out
Every marketing message needs a visible, working exit. If unsubscribing is easy, unhappy readers quietly leave your list. If it is hard, they use the two buttons WhatsApp gives them instead: Block and Report, and those are the exact signals that sink your quality rating.
The accounting is lopsided in your favour. An unsubscribe costs you one contact who was never going to buy. A block feeds the 7-day rating that throttles your reach to everyone else. Routing unhappy readers to an unsubscribe button is not politeness; it is armour.
Here is what that looks like in practice. Businesses and names below are fictional.
Illustrative template with a fictional business.
And the version that quietly destroys lines: arguing with a stop. One more message after someone asks out is enough to turn a lost subscriber into a spam report.
Illustrative conversation with a fictional business. The stop request must end the sequence, instantly and silently.
For reminder sequences, we add a softer exit alongside the hard one: a Pause reminders button. It lets people mute a sequence without leaving your list, which protects both the relationship and the rating.
Rule 3: watch the ratio, not just the volume
Meta's systems watch the flow around your number: outgoing messages on one side, and reactions (delivered, read, blocked, reported) on the other. What sinks numbers is not volume itself but volume whose reactions run negative: a spike of sends to people who did not expect them.
Compare two months of identical volume. Ten thousand marketing blasts to lukewarm contacts will harvest blocks at a rate that flags the number. Ten thousand order confirmations, booking reminders and receipts will harvest almost none, because every message answers something the customer just did.
Same volume, opposite fates. The ratio of wanted to unwanted messages is the lever, and it is entirely under your control before you press send.
The Zelix way: utility messages first
Our standing rule for every new client line: set up utility messages before any marketing goes out. Utility messages are the transactional updates a customer expects because of something they just did: an order confirmation, a payment receipt, a booking reminder, a delivery update (utility templates, as Meta categorises them).
Think about the last time you bought something online and the store sent an order confirmation. Did you consider blocking it? Nobody does. Nobody blocks a receipt, because losing it costs the customer more than reading it.
Now scale that up. If a thousand people order from a store and each gets a confirmation on WhatsApp, the block count is effectively zero. The messages are wanted, read quickly, and often replied to, which is exactly the reaction profile that builds a number's standing.
That is the whole strategy. Months of utility traffic give Meta a long, clean record of your number sending things people demonstrably want. You are not gaming the system; you are feeding it the truth that your messages are welcome.
Illustrative template with a fictional business.
Illustrative template with a fictional business.
One trap to avoid while writing these: keep utility templates plain. A promotional line or a salesy flourish inside a utility template can get it re-classified as marketing by Meta's review, which changes its price and its scrutiny. The receipt is the message; the upsell belongs elsewhere.
A good history is a shock absorber
Meta's enforcement reads your number the way a bank reads a borrower: history counts. A line that has sent months of high-volume, low-complaint utility traffic has proven its messages are wanted, and that record cushions the occasional bad day.
Picture the failure case every operator eventually meets: a marketing broadcast goes out and someone forgot the unsubscribe button, and a few hundred people block it. On a seasoned line, that is a bad week: the rating dips, you fix the template, and the history pulls the rating back.
The same mistake on a fresh number with no outgoing history is often fatal, because the blocks are the only signal Meta has ever seen from it. This is why we refuse to launch marketing on a cold line, ever. Warm it up with utility traffic first, then earn the right to promote.
Ask permission at every door
Consent is collected at the moments a customer hands you their number: checkout pages, enquiry forms, booking flows. We put a WhatsApp opt-in at every one of those doors, worded plainly, stored with a timestamp, and separated into two different asks.
The first ask covers order and booking updates, and almost everyone ticks it because the updates help them. The second covers offers and news, and fewer people tick it, which is the point: the marketing list you end up with actually wants marketing.
A purchase by itself is consent for messages about that purchase, not for promotions. Keeping the two asks separate is also what privacy regimes like the PDPA and do-not-call rules expect (DNC registry), so the same checkbox that protects your line keeps you legal.
Illustrative checkout consent block. The utility ask and the marketing ask stay separate.
Send at human hours, at a human pace
Timing decides how a message is received before a single word is read. A promotion at 7am reads as spam from a business; the same promotion mid-morning reads as a note from a shop you know. Blocks follow annoyance, so cadence discipline is quality protection.
Our house rules, in the contact's local time zone:
- Quiet hours are absolute. Nothing scheduled between 9pm and 9am. Customer-triggered utility (a receipt for a purchase they just made) is the only exception.
- Marketing sits inside 9am to 8pm, aimed at mid-morning or early evening.
- Roughly two marketing broadcasts per contact per month, at most. Past that, block rates climb no matter how good the offer is.
- Follow-ups stop early. A nudge at +24 hours, one more at +72, then silence. Persistence past two unanswered messages converts almost nothing and reports plenty.
Anatomy of a throttling: two weeks at a restaurant
Here is how the escalation plays out in real life, compressed from patterns we guard against. The business is fictional; the sequence is not.
- Day 1. A restaurant wants a New Year's Eve push. Staff export 1,800 delivery numbers from two years back and blast a promo. No opt-in gate, no unsubscribe exclusion.
- Day 2. Ninety people tap Block; dozens report "didn't sign up". The 7-day quality rating flips from green to red.
- Day 3. The promo template is paused for negative feedback, and the daily conversation allowance steps down a tier, mid-campaign.
- Day 7. Legitimate sends now queue against a smaller allowance. Same-day booking confirmations still flow, but the day-before reminder templates start hitting the ceiling. The spam campaign is now blocking the restaurant's own receipts.
- The recovery. All marketing stops, every opt-out is honoured, and sends resume only to opted-in segments. Quality crawls back over the following weeks, and the allowance climbs only after the rating recovers.
That account ended two escalation steps from a lock. One list, one send, weeks of damage. Every rule in this guide exists to make that fortnight impossible.
The line-quality checklist
The whole playbook on one screen. Run it before every campaign, and put it in front of anyone new who touches the number.
Protects the number
- Utility messages live before any marketing
- Every contact opted in, with a record of when and how
- Unsubscribe button on every marketing template
- Pause option on every reminder sequence
- Stops honoured instantly, no parting shot
- Marketing capped near two sends per contact per month
- Broadcasts inside 9am to 8pm local, never quiet hours
- New numbers warmed up with small opted-in batches
Burns the number
- Bought, scraped or exported cold lists
- Marketing as the first traffic a new number sends
- Promotions with no visible way out
- Replying to a stop request with an offer
- Volume spikes to lukewarm audiences
- Promotional copy hidden inside utility templates
- Late-night or early-morning blasts
- Opening a replacement account during an appeal
Questions we hear a lot
Is the quality rating per template or per number?
Both. The number carries the 7-day rating, and each template carries its own score. A single template can be paused for bad feedback while the number stays healthy - treat that as the early warning it is.
Does getting a new number reset a bad rating?
It restarts the problem, not the solution. A fresh number starts at the lowest tier with zero history, so the same behaviour fails faster. And a replacement account opened during an appeal is treated as evasion. Fix the list and cadence instead.
What ratio of utility to marketing keeps a number safe?
Meta publishes no official ratio. Our working rule: the bulk of your outgoing volume should be utility people expect, with marketing capped near two broadcasts per contact per month, always to opted-in segments.
Do unsubscribes hurt my quality rating?
No. An unsubscribe costs you one contact. A block or report feeds the rating that throttles the whole number. The unsubscribe button exists to catch people before they reach for the block button.
Can I message customers who bought but never opted in?
About their purchase, yes: confirmations, receipts, delivery updates. For promotions, no - collect a separate, recorded marketing opt-in first, and honour opt-outs instantly.
What actually happens before a ban?
Escalation in steps: a template is paused, then all messaging is blocked, then the account locks, then the number is removed. A ban is a pattern of ignored warnings, and every step is a chance to stop and recover.
Final words
None of this is a trick, and that is the point. Meta's systems are built to answer one question: do people want what this number sends? Buy no lists, give every reader an exit, keep the wanted-to-unwanted ratio overwhelming, and let utility traffic build your history before marketing spends it.
Run a line that way and the platform works for you: the allowance grows, campaigns deliver, and one bad day stays one bad day. If broadcasts are part of your plans, our follow-ups setup guide covers the template mechanics.