The WhatsApp anti-ban playbook: operating a quality rating

Most advice about avoiding WhatsApp bans is a list of things not to do. That is necessary and it is not a playbook, because it gives nobody a routine to follow on an ordinary Tuesday when nothing is wrong.

The operational version treats the quality rating as a number with an owner, a baseline, thresholds and a defined response — the same way a team would treat error rate or uptime. That is unglamorous and it is what actually prevents the situation the lists warn about.

What we'll cover

• Treat it as an operated metric
• Warming up a new number
• What to watch, and how often
• Building it with CodeWords
• Thresholds and the response to each
• The pre-send checklist
• Running a campaign safely
• The weekly review
• Structuring accounts so one mistake is contained
• Making it survive
• Limits worth knowing about
• What to do first
• Frequently asked questions

Treat it as an operated metric

Name an owner, an individual, who looks at it as part of a routine.

Establish the baseline, since a rating you have never looked at gives you no sense of normal.

Define the thresholds and the response to each before anything is wrong.

Put it on a dashboard somebody already opens, rather than in a console they visit when worried.

Review it weekly even when nothing has changed, which is what builds the sense of normal.

It is a leading indicator, which is the whole reason to operate it — restrictions follow rating declines rather than arriving unannounced.

Warming up a new number

Start low regardless of what you intend to send eventually.

Begin with inbound and replies, which carry the least risk and the most positive signal.

Then transactional messages people are expecting — confirmations, updates, reminders they asked for.

Then requested notifications, and only then anything promotional.

Increase gradually rather than in steps that look like a switch being flipped.

Give it weeks, not days. A number that goes from zero to a campaign is the pattern that attracts attention.

What to watch, and how often

Quality rating, daily during any campaign and weekly otherwise.

Messaging limit and consumption, so you know your headroom.

Block rate where visible, which is the underlying driver of the rating.

Template performance individually, since the aggregate hides the one doing damage.

Opt-out requests, which are both an obligation and an early signal.

Response rate, because a message nobody replies to is a message nobody wanted.

Building it with CodeWords

CodeWords connects to more than 3,000 integrations, and the connection is made once and reused.

1. Open CodeWords and start a new automation.
2. Describe what should happen in plain language to Cody, the automation builder: what is monitored, what thresholds apply, and who is alerted.
3. Authorize the connections to your WhatsApp Business Platform account and your alerting channel.
4. Describe the exceptions: a rating change, a limit approaching, a template performing badly, an opt-out.
5. Run it in report-only mode for a fortnight to establish the baseline before setting thresholds.

You describe the outcome; Cody builds it, connects it, and deploys it. The free plan covers light use, with Pro at $39 per month and Business at $100 per month as usage grows; details are on the pricing page.

Thresholds and the response to each

Green and stable: proceed, review weekly.

Any decline from green: stop the newest campaign, identify what changed in the preceding days, do not start anything new.

The middle state: halt all promotional sending, continue transactional and replies only, review every active template.

The lowest state: stop outbound entirely, expect a limit reduction, and begin the recovery period rather than appealing first.

Recovering: resume slowly and with wanted messages, not with the campaign that caused it.

Write these down and give somebody the authority to act on them without a meeting, since the response has to be same-day to matter.

The pre-send checklist

Has this audience opted in for this kind of message, specifically, and can you show it?

Is the template category correct for the content, which is a common and avoidable finding?

Would you want this message? The honest answer predicts the block rate better than any tool.

Is there an opt-out and does it work?

Is the timing reasonable in the recipient's timezone?

Is the volume within headroom, including the portfolio allowance shared with other numbers?

Running a campaign safely

Send to your most engaged segment first, which both performs better and protects the rating.

Stagger across days rather than in one send.

Check the rating between batches, which is only possible if you staggered.

Set a stop condition in advance — a block rate, a rating change — and honour it.

Have somebody watching during the send rather than reviewing afterwards.

Stop on the signal, since a campaign paused halfway costs less than a restriction.

The weekly review

The rating trend, not just the current value.

Templates by engagement, retiring the ones nobody responds to.

Opt-outs by source, which tells you which list or form is producing unhappy recipients.

Limit consumption against the tier, and whether advancement is plausible.

Anything Meta sent you, which is easy to miss in a shared inbox.

Fifteen minutes, weekly, which is the whole cost of this.

Structuring accounts so one mistake is contained

Most businesses run everything through one number under one portfolio, which means a marketing campaign and a delivery notification share a fate.

Separate by consequence, not by team. The messages that must arrive — order updates, appointment confirmations, authentication — are a different risk class from promotional sending.

A separate number for transactional messaging means a marketing rating problem does not stop a customer receiving their booking confirmation.

Understand the shared allowance. Since October 2025 numbers in one portfolio share a messaging limit, so separation by number does not separate capacity.

Separate portfolios genuinely separate, and they need their own verification — this is a legitimate structure for genuinely separate business units rather than a device for evading limits.

Decide before you scale, because restructuring once customers have saved a number is the expensive version.

Document which number does what, since the reason for the split is exactly what gets forgotten.

Making it survive

Alert on rating change to a person, not a channel, with the authority to stop a send.

Keep opt-in evidence with wording and date, retrievable per recipient.

Suppress opt-outs across every automation, immediately and permanently.

Do not let campaign tooling bypass the checks, which is how a marketing send goes out without the pre-send list.

Report the outcome weekly: rating, limit, sends, opt-outs, and templates retired.

Limits worth knowing about

The rating is a summary and you do not see the calculation behind it.

It lags slightly, so a bad send is visible after some of the damage is done.

Recovery periods are not published, so plan for longer than you hope.

Portfolio-level limits mean another team's campaign can consume your headroom.

None of this substitutes for wanting the message, which is the actual variable.

What to do first

Establish your baseline: rating, tier, consumption, and the templates currently active — recorded, with an owner's name against them. It takes half an hour, most teams running this channel cannot produce it, and every threshold and response in this article depends on knowing what normal looks like for you.

Two habits make the difference. Give one named person the authority to stop a send without a meeting, since the response to a rating decline only works if it is same-day. And send to your most engaged segment first in every campaign, because it protects the rating while producing better results, which is the rare case where the safe option is also the effective one.

Frequently asked questions

How should quality rating be managed?

As an operated metric with a named owner, an established baseline, written thresholds and a defined response to each — reviewed weekly even when nothing is wrong, which is what builds a sense of normal.

How do I warm up a new number?

Start with inbound and replies, then transactional messages people expect, then requested notifications, and only then anything promotional. Increase gradually over weeks — a number going from zero to a campaign is the pattern that attracts attention.

What should I watch?

Quality rating daily during campaigns and weekly otherwise, limit consumption, block rate where visible, template performance individually rather than in aggregate, opt-out requests, and response rate.

What is the response to a rating decline?

Stop the newest campaign rather than reducing it, identify what changed in the preceding days, and start nothing new. At the middle state, halt promotional sending entirely and continue only transactional messages and replies.

What belongs on a pre-send checklist?

Whether this audience opted in for this specific kind of message and whether you can show it, whether the template category matches the content, whether the opt-out works, the timing in the recipient's timezone, and the headroom including the shared portfolio allowance.

How should a campaign be run?

To the most engaged segment first, staggered across days, with the rating checked between batches, a stop condition agreed in advance, and somebody watching during the send rather than reviewing afterwards.

What is the single best predictor of a block?

Whether you would want the message. It predicts block rate better than any tooling, and it is the check most often skipped because it is uncomfortable rather than difficult.

How much time does this take?

About fifteen minutes weekly once the baseline exists, plus attention during campaigns. That is the whole operational cost, against a restriction that takes weeks to recover from.

Should transactional and marketing messages share a number?

Separating them by consequence means a marketing quality problem does not stop customers receiving booking confirmations or order updates. Decide before you scale, since restructuring after customers have saved a number is the expensive version.

Does using several numbers give me more capacity?

Not within one portfolio — since October 2025 the messaging limit is shared across all numbers in it. Separate numbers separate risk, not allowance.

Who should own the quality rating?

A named individual with the authority to stop a send without convening anybody. A rating decline needs a same-day response, and a metric owned by a team is checked by nobody.

What does a healthy baseline look like?

Whatever yours is when nothing is wrong — which is why establishing it before you need it matters. A rating you have never looked at gives you no way to recognise a decline until a restriction arrives.

Related reading

• Why WhatsApp bots get banned
• WhatsApp messaging limits explained
• WhatsApp automation mistakes that get numbers flagged