It is early on a Sunday and the marketing coordinator at an aesthetics clinic in Dubai is refreshing the WhatsApp Manager tab for the fourth time. On Friday afternoon she pushed a single large offer send to a contact list assembled from old walk-in records, an event sign-up sheet and a file handed over by a former agency. By Saturday evening the indicator beside the phone number had turned yellow. This morning two templates carry a paused status, the clinic's appointment reminders are not going out, and the front desk has gone back to phoning people one at a time.
That is the whole problem in one weekend. A quality rating is not a vanity metric. It is the gate that decides whether your reminders, your confirmations and your follow-ups reach anyone tomorrow.
What the rating is actually measuring
A WhatsApp quality rating moves from yellow to green when the negative feedback Meta collected over the trailing seven days ages out of the window and is replaced by clean sending days. Meta scores each business phone number on how recipients reacted to recent messages, and according to Webex Connect's platform documentation that rating covers the past seven days weighted by recency, built from user feedback signals like blocks, reports and the reasons people give when they block a business. There is no appeal form and no support ticket that resets it. You change the inputs, you wait out the window, and the colour follows.
Two things follow from that. The first is that recency weighting cuts both ways: a bad send yesterday carries more weight than one last Tuesday, and a clean day today counts for more than a clean day a week back. The second is that the number-level rating and the template-level rating are separate scores. Meta's own template quality documentation defines yellow at the template level as medium quality caused by negative feedback from multiple users or low read rates, with a warning that the template may soon be paused or disabled. Note also that AiSensy's help centre describes the API quality rating as reflecting message quality over the past 24 hours, which is a useful reminder that different surfaces in the ecosystem report on different windows. The practical reading: the score is short-memoried, and short memory is the thing working in your favour.
What each colour costs you
The three states are defined the same way everywhere. Green is high quality. Yellow is medium. Red is low. Kanal's explainer on quality scoring describes green as a number in good standing with low block and report rates, where Meta is happy to let the messaging limit grow, and yellow as the state where elevated negative signals have been detected, often after a single bad broadcast. Meta's template documentation is blunt about red: low quality, with templates in that state at risk of being paused.
Sitting underneath the colour is your messaging tier. WUSeller's breakdown of the tier structure lists the ladder as 250 unique users per day at Tier 1, then 2,000, 10,000, 100,000, and unlimited at Tier 5, and states that WhatsApp assigns the tier automatically, increasing it only when it trusts your behaviour, with trust built over time through consistent behaviour. You do not apply. You behave, and the ladder moves.
The seven signals Meta weighs
Before you fix anything, know what is being read. Pulling the documented signals together, these are the seven inputs that decide your colour:
- Blocks. Named explicitly in Webex Connect's documentation as a user feedback signal feeding the rating.
- Reports. Also named explicitly, and heavier than a block because it carries an accusation, not just a dismissal.
- Block reasons. The documentation specifies that the reasons users select when they block a business form part of the signal set. Not all blocks are read identically.
- Read rates. Meta's template quality page names low read rates alongside negative feedback as a cause of yellow template status.
- Recency of all of the above. The seven-day window is weighted by recency, so timing is itself a signal.
- Per-template feedback. Templates carry their own green, yellow and red status independent of the number, driven by feedback from multiple users on that specific template.
- Behavioural consistency over time. WUSeller describes tier increases as a function of WhatsApp trusting your behaviour, built through consistency, which is a longer-arc signal than the seven-day feedback window.
Everything else you will read about quality ratings is a proxy for one of those seven. Frequency matters because it produces blocks. List quality matters because unpermissioned recipients report. Copy matters because it drives read rate. Work backwards from the signal, not forwards from the tactic.
Reading the instruments before you touch the controls
Air traffic control does not vector an aircraft off a bad approach by guessing. The controller reads position, altitude and closure rate first, then issues one instruction, then watches what it did. Diagnosis, single correction, confirmation. Most of the yellow-rating panics we get called into start with the opposite: the operator has already deleted templates, changed the display name and started a new send to a supposedly cleaner list, which adds three fresh variables to an already noisy seven-day window and makes it impossible to tell which change did what.
So read first. Open WhatsApp Manager, find the phone number, and note four things: the current colour, the current messaging tier, the per-template quality status, and any block or report reasons your provider surfaces. Spur's help centre shows where the messaging limit and quality rating sit side by side in the interface, which is the view to keep open while you work.
Then answer one question honestly: which send caused this? In almost every case we investigate, the answer is one campaign, one list, one day. Kanal makes the same observation, that a single bad broadcast is frequently enough to tip a number from green into yellow. Find that send before you change anything else.
The corrections, in the order they should be made
Stopping the sends you cannot evidence consent for
This is the correction that does most of the work, and the one operators resist hardest because it shrinks the addressable list. Best practice, and the only version of this we will install, is that a contact receives marketing templates only where you can point to the moment they agreed to receive WhatsApp messages from your business specifically. Contacts inherited in an agency handover file rarely clear that bar, and in our experience they are the usual source of the reports that turned the number yellow.
Segment the database by consent provenance: people who messaged you first, people who gave explicit consent naming your business, and people you cannot evidence. Send to the first two. Stop sending to the third. If you want to re-permission that third group, do it through a channel they did consent to. We have written separately on the exact wording that stands up, so we will not repeat it here.
Bringing frequency back to what a recipient would predict
Blocks cluster around surprise. Someone who booked a single treatment months ago and receives a run of promotional templates inside a fortnight does not report you because the offer is weak. They report you because the cadence feels unrelated to any relationship they think they have with you. We cap promotional templates at a rhythm a reasonable customer would anticipate, and we keep utility messages (confirmations, reminders, receipts) architecturally separate from marketing, so that a marketing problem never takes the operational messages down with it.
Rewriting the templates nobody opens
Read rate is a scored signal, not just a marketing KPI, because Meta's template documentation names low read rates as a cause of yellow status in its own right. A template with a generic opener, no personalisation and no obvious reason for arriving gets left unopened, and unopened at volume reads to the algorithm as irrelevance. Rewrite the first line so it names the specific thing the recipient did or booked. Delete the template that opens with a greeting and reaches the point in sentence three.
Pausing the template, not the channel
When a number goes yellow, the reflex is to stop everything. That is usually the wrong move, because a total stop also stops you generating the clean days that rebuild the score. Pause the specific templates showing yellow or red. Keep utility messages flowing to people who are actively in conversation with you. A number that goes silent produces no negative signals, but it also produces no evidence of good behaviour, and the window is still ticking either way.
Making a reply possible, and answering it
Two-way conversation is the strongest counterweight to blocks. A template that invites a reply and gets one produces engagement; a template arriving from a number nobody can answer offers the recipient exactly one interaction, and it is the block button. Quick reply buttons help. A human who answers within business hours helps more. This is where our second house position stops being a philosophy and becomes an operating requirement: the automation exists to handle acknowledgement and routing so the person at the desk has time to answer the messages that need a person.
Aligning the business details behind the number
Trust signals sit outside the message stream too. Chatarmin's guide to Meta business verification lists mismatches between Business Manager, business registration and website imprint (differing addresses, differing phone numbers) and incomplete business verification as common problems. Align them and complete the verification. A recipient who sees a display name matching the clinic they actually visited is less likely to reach for report than one who sees an unfamiliar trade name.
Cleaning the database before the next send
Most implementations we are asked to rescue fail on data, not on the AI or the automation layer, because a system fed inconsistent records acts confidently on the wrong ones. Duplicate contacts, numbers stored in three different formats, a last-visit field populated by two systems that disagree: send into that and you will message the same person twice in an hour, message someone about a service they never bought, and attempt delivery to a landline. Every one of those is a negative signal in waiting. Deduplicate, normalise to E.164, and remove anything you cannot trace to a source. The platform mechanics behind all of this are covered in our WhatsApp platform reference.
The same reminder run, done two ways
Take one workflow, appointment reminders for tomorrow, and watch it twice. The clinic below is a composite of projects we have worked on, and the counts are illustrative rather than measured.
Manually, it runs like this. Late afternoon, the coordinator exports tomorrow's appointment list from the practice management system into a spreadsheet and sorts it by time. She copies each number into WhatsApp Web, pastes a reminder she keeps in a notes file, edits the name and the time by hand, and sends. Partway down the list she pastes the wrong time into a reminder. Several numbers fail because they were entered with a leading zero and no country code. A couple of patients cancelled after the export ran, so they get reminders for appointments that no longer exist, and one of them replies irritated. The patients who reply to confirm land in a shared inbox nobody opens until the following morning, so nobody is acknowledged. And when she runs out of time, the remainder go out as one identical copy-paste to a broadcast list. That last move is the one that generates block reports.
Automated, the same run looks different at every joint. The system reads tomorrow's confirmed appointments live from the practice management system rather than from an export, so a cancellation made minutes earlier is already excluded. Numbers are normalised to E.164 before anything is attempted, and records without a valid mobile are flagged to the desk instead of being sent into the void. Each patient receives an approved utility template with their own name, treatment and time merged in, delivered as an individual conversation rather than to a list. The template carries two quick reply buttons, confirm and reschedule. A confirm writes straight back to the appointment record and returns an instant acknowledgement. A reschedule opens a conversation routed to the coordinator with the appointment already attached, so her late afternoon goes to the handful of patients who need a human instead of the rest who need a timestamp.
The delta that matters for quality rating is not the time saved. It is that the automated run sends nothing to invalid numbers, nothing about cancelled appointments, nothing identical in bulk, and produces a steady stream of inbound replies that reads as healthy two-way engagement. The manual run produces the exact signal profile that turns a number yellow.
How recovery tends to run
Because the rating window is the trailing seven days weighted by recency, the mechanics point to a clean week as the shape of a recovery, with the most recent days pulling hardest. We have watched numbers flip back quickly when the cause was one identifiable send that stopped immediately, and we have watched numbers sit yellow far longer when the operator kept sending to the same unevidenced list while waiting for the colour to change. Those are our observations from the work, not published figures, and no source we trust publishes a guaranteed recovery time. Anyone who quotes you one is guessing.
Red is different in character. Red is the state where restriction becomes a live possibility, and the right response is to stop marketing templates entirely, keep only utility messages inside active conversations, and treat recovery as a project rather than a weekend fix. If the account has also drawn a review, that is a separate process with its own triggers, which we cover in our piece on what triggers account reviews.
The move that makes things worse
Do not buy a second number and move the same list onto it. A fresh number starts at the bottom of the ladder, which per WUSeller's tier structure means 250 unique users per day, and the pattern of a brand-new number immediately pushing high volume to a cold list is precisely the behaviour the trust-building tier system is built to notice. You would be trading a yellow number with real capacity for a green number with almost none.
Verifying the fix held
A controller does not treat a handoff as complete because the pilot read back the instruction. The aircraft has to be on the new heading, confirmed on the scope, before attention moves on. Same discipline here: the colour turning green is the read-back, not the confirmation.
Check these, in this order, once you have a full clean week behind you:
- Number-level rating in WhatsApp Manager shows green, and has held green across two checks several days apart rather than flickering.
- Every active template shows green individually, with no paused or disabled status sitting behind a healthy number-level score.
- Messaging tier is stable or climbing. WUSeller notes tiers rise only as WhatsApp trusts consistent behaviour, so a flat tier alongside green is normal while volume is still modest.
- Read rate per template is trending up rather than flat. Read rate is the leading indicator; the colour is the lagging one.
- Inbound replies relative to outbound sends. Send a hundred templates and receive nothing back and you are still producing a one-way pattern, whatever the dashboard says today.
Then build one habit: check the rating on the same day every week, not only when something breaks. Learnmind is an AI customer-communication consultancy in Dubai, and the thing we find most often when a client calls in a panic is that nobody had looked at the number in months. Yellow is a warning, and it is designed to be acted on while sending still works.
Quick answers
How long does it take for a WhatsApp quality rating to go from yellow to green?
There is no published recovery time, but because Meta evaluates the trailing seven days weighted by recency, a full week of clean sending is the natural shape of a recovery. Numbers where the offending send has not actually stopped will stay yellow regardless of how long you wait.
Can I contact Meta to reset a yellow quality rating?
No. There is no appeal or reset process for a WhatsApp quality rating, because the score is recalculated automatically from recipient feedback signals like blocks and reports. The only way to change it is to change the list, the frequency and the messages that produced it.
Does a yellow quality rating stop my WhatsApp messages from sending?
No. Meta's template quality documentation states that templates rated yellow can still be sent. What yellow signals is that the template has drawn negative feedback from multiple users or low read rates and may soon be paused or disabled if the pattern continues.
We will audit your WhatsApp number, trace the send that turned it yellow, and rebuild your reminder and follow-up templates so the rating holds green as your volume grows.
