This article is for marketing operations leads and CRM owners who run outbound WhatsApp for a hospital, a hospital group or a multi-clinic chain and want to scale volume without damaging the number. It assumes you already have a verified account and approved templates. For the basic strategy, read our WhatsApp marketing guide for healthcare first. This post goes into the mechanism: tiers, quality signals, ramp plans and what to do when quality drops.
A healthcare number has one advantage and one weakness. The advantage is that many of its messages are wanted: reminders, report alerts, confirmations. The weakness is that a single bad campaign to an unengaged list can produce blocks and reports that affect the delivery of those useful messages. In a hospital, a throttled number is not a marketing inconvenience. It can delay appointment reminders.
We rely on Meta's current developer documentation. Where Meta does not publish a number or state, we say so and describe how to measure instead.
How messaging limits work
According to Meta's messaging limits documentation, a limit is the maximum number of unique WhatsApp users a business can message with business-initiated messages in a moving 24-hour period. The documented levels are 250 for new portfolios, then 2,000, 10,000, 100,000 and unlimited.
Several details change how you plan:
- Limits belong to the business portfolio. They are shared by every phone number in the portfolio. If a hospital group runs separate numbers for the main hospital, the cancer centre and the diagnostics arm under one portfolio, one number's campaign can consume capacity the others need.
- The count is unique recipients, outside customer service windows. Sending three messages to one patient counts once. Replies inside an open 24-hour window do not count against the limit.
- It is a moving period, not a calendar day. A burst at 10 am can still be limiting at 9 am the next day.
How to move up a tier
To reach 2,000, Meta lists three scaling paths: verify your business, have your partner verify it if a partner onboarded you, or send 2,000 delivered messages outside customer service windows to unique numbers within a 30-day moving period using high-quality templates. Meta then analyses message quality before approving.
Above 2,000, scaling is automatic and needs both of the following: high-quality messages across all numbers and templates, and use of at least half the current limit in the last seven days. When both hold, the limit rises one level within six hours.
You can read your current limit in WhatsApp Manager under account tools, or through the API field whatsapp_business_manager_messaging_limit. The older messaging_limit_tier field is deprecated, so check integrations that still read it.
The implication for a hospital is that you cannot ramp by volume alone. A large, low-engagement send does not move you up and can set you back. Quality is the gate.
What "quality" is built from
Meta does not publish a formula. The documentation points to three kinds of signal that you can observe or influence.
Recipient feedback. Meta's opt-in guidance notes that people can block or report a business, can stop marketing messages from an individual business through WhatsApp's user preferences, and that sustained low quality can lead to rate limiting.
Template quality. Meta's template quality documentation describes four ratings reported by the API: green for high quality, yellow for medium, red for low, and unknown while no feedback or read-rate data exists yet. Yellow and red arise from negative feedback from multiple users or low read rates. Both can still be sent, but a template at risk can be paused or disabled, and a template that leaves the approved status cannot be sent.
Read behaviour. Low read rates count against a template. A message that recipients do not open is a signal that they did not want it or did not expect it.
Meta does not hand you a block count for each campaign. You infer problems from proxies, which is why monitoring needs to be designed rather than assumed.
Build the monitoring layer
Track these at the template and segment level, daily during a ramp and weekly otherwise.
| Signal | Source | What it tells you |
|---|---|---|
| Template quality score | Template API field quality_score, or Manage templates panel | Direct Meta rating per template |
| Template status changes | message_template_status_update webhook | Pause or disable events |
| Delivery rate | Message status webhooks | Wrong numbers, unreachable users |
| Read rate | Message status webhooks | Interest and expectation |
| Opt-out replies | Inbound text such as STOP | Direct dissatisfaction |
| "Wrong number" or "who is this" replies | Inbound text classification | Capture-process failures |
| Messaging limit field | Business portfolio API | Tier movement |
Set alert owners. Someone must read the pause webhook the same day. A template that has quietly turned yellow for two weeks is a campaign someone sent without looking.
Treat inbound objections as a metric
Tag inbound messages that contain opt-out phrases, "wrong number", "I did not ask for this" and complaints. Count them per thousand delivered messages by template and by source of the number. This is the closest practical proxy for block behaviour you can build yourself, because every blocked recipient was first someone who did not want the message. If one source, such as an old camp list, produces most of the objections, you have found the problem.
Segment before you send
Quality is controlled mostly before the send button. Segment on four axes:
- Consent purpose. Only patients who agreed to promotional messages receive marketing templates. Our opt-in management service covers the consent side.
- Recency of relationship. A patient seen last month is a different audience from one seen in 2021. Send to the recent group first.
- Engagement history. Contacts who replied, clicked or booked via WhatsApp go ahead of those who have never responded.
- Specialty relevance. A cardiology awareness message to a paediatric list is the kind of mismatch that generates blocks.
Exclude numbers that bounced, patients flagged as deceased in the records, and people who withdrew consent. Excluding them costs nothing and protects the portfolio.




