Most multi-branch businesses in the UAE need one WhatsApp number, not one per branch. Meta caps a new business portfolio at two registered numbers and raises it to 20 once your business is verified, so the technical ceiling is rarely what stops anyone. The cost of extra numbers is what stops them. From 1 October 2026 Meta gives every business phone number 1,000 free service messages a month and charges utility rates after that, which sounds like an argument for more numbers until you price it: that allowance is worth roughly AED 30 to 60 a month at UAE utility rates, and hosting an extra number with a provider costs AED 40 to 75 a month. The free tier does not pay for the number that carries it.
For a salon with three locations or a clinic group with nine, the real question is where the routing decision gets made and how fast the customer finds out which branch is answering, not how many numbers to buy.
In short: WhatsApp branch routing is the logic that sends an inbound WhatsApp enquiry to the right location's team, calendar and CRM pipeline. It can sit in four places: the entry point the customer tapped (a per-branch link, QR code or ad), the phone number they messaged, the conversation itself (an AI agent that asks or infers the location), or the CRM record that already knows which branch owns this customer. One WhatsApp Business Account can hold 20 numbers after Meta business verification, and each number carries its own approved display name, so "Brand Al Barsha" and "Brand JLT" can live under one account.
What changes on 1 October 2026, and what it costs per branch
Until now, anything you sent inside the 24-hour customer service window was free. Template messages were billed by category and everything else rode along. From 1 October 2026 Meta bills service messages too, at the utility rate, after a free allowance of 1,000 delivered service messages per business phone number per month. The allowance resets monthly and does not roll over.
For a single-location business this is a rounding error. For a group with branches it changes the arithmetic of how many numbers you run, because the allowance is per number rather than per account. Here is that arithmetic for a four-branch clinic group doing roughly 6,000 service messages a month across all locations, using the UAE utility band of about AED 0.03 to 0.06 per message and a typical provider hosting fee of AED 40 to 75 per number.
- One number for all four branches. 1,000 messages free, 5,000 chargeable at AED 150 to 300, one hosting fee at AED 40 to 75. Monthly total AED 190 to 375.
- One number per branch, 1,500 messages each. 4,000 messages free, 2,000 chargeable at AED 60 to 120, four hosting fees at AED 160 to 300. Monthly total AED 220 to 420.
Four numbers cost AED 30 to 45 a month more than one, not less. The extra free allowances are real, and they are smaller than the hosting fees that unlock them. It only tilts the other way if every branch number independently clears 1,000 service messages a month and your provider bundles hosting rather than charging per number.
That settles one thing: pick your number count for operational reasons. Nobody should be adding WhatsApp numbers to save money on Meta's bill.
Four ways to route WhatsApp across branches
- Option A, one number with routing in the conversation. The customer sees one brand number everywhere. The AI agent asks or infers the location and assigns it in the CRM. Fits 2 to 15 branches in one country under one management team.
- Option B, one account with a number per branch. The customer sees a different number on each location page and Maps listing. The routing is the number itself, read by the webhook or the inbox. Fits branches with their own managers and their own calendars.
- Option C, a separate account per branch. Different numbers and fully separate accounts, with the split at account level. Fits franchisees who own their own legal entity and want their own data.
- Option D, hybrid. A few numbers split by emirate or by language, one per group of branches, with everything finer handled inside the conversation. Fits groups spanning Dubai, Abu Dhabi and Sharjah, or an Arabic and English split.
Option A is what most UAE SMBs should run and rarely do. One number means one display name to get approved, one verification path, one template library, one set of quality ratings to protect, and one place where every conversation is logged. Branch selection happens inside the chat, which costs nothing and can be changed on a Tuesday afternoon without touching Meta.
Option B is the one people assume they need. It is a genuine fit when branch managers hold their own P&L and their own calendar, because the number becomes the boundary of accountability. It also carries real weight in local search: a Google Business Profile for each location with its own WhatsApp link routes traffic before a single message exists. The cost is administrative. Each number needs its own display name approved, its own quality rating watched, and its own phone line or landline that can pass verification.
Option C is heavier than it looks in the UAE, and it is where most multi-branch plans quietly fail. Meta business verification asks for a legal entity document, and a UAE branch licence usually carries the same parent legal name and the same trade licence number as head office. A branch of a Dubai mainland company is not a separate legal person; a franchisee with their own licence usually is. If you are promising a franchisee their own account, confirm the entity structure with your provider before you promise it, and expect freezone, mainland and different-emirate licences to behave differently. We cover what Meta actually asks for in the UAE verification walkthrough.
Option D is what groups that cross emirates usually land on. One number for Dubai, one for Abu Dhabi, one for the Arabic-first line. The broad split is structural and stable, and everything finer than that happens inside the conversation, where it can be changed without a Meta review.
Route at the entry point, before the message exists
The cheapest routing layer in the entire stack is the link the customer tapped, and almost nobody uses it properly.
A wa.me link can carry prefilled text. Put a branch identifier in it, publish a different link on each location page, each Google Business Profile, each printed QR code at each reception desk, and the first message that lands already says which branch the customer came from. No menu, no question, no lost second. A customer who scanned the code on the counter in Al Quoz never has to be asked where they are standing.
Click-to-WhatsApp ads work the same way. Our own paid chats arrive with the ad's autotext as the opening line, which is how we separate a paid enquiry from an organic one without a tracking pixel doing anything clever. If you run one ad set per branch with a distinct opening line, attribution and routing arrive in the same string.
This matters more than any inbox feature, because entry-point routing is the only kind that is invisible to the customer. Every other layer costs the customer something, even if only a tap.
What breaks in practice
We run a WhatsApp agent on our own business. It qualifies the enquiry, prices the job, books the slot, sends the payment link, and hands to a human when the conversation stops being a booking and starts being a negotiation. Most of what follows we learned by getting it wrong first.
A branch menu kills the first message. The instinct is to open with "Reply 1 for Al Barsha, 2 for JLT, 3 for Marina, 4 for Deira, 5 for Sharjah". Every option you add is a chance for the customer to close the chat instead of choosing. People message a business because they have a question, and a numbered list is a form. If you must ask, ask after you have answered something, and ask with two or three named areas rather than a full directory. Better: infer it. A customer asking about a service only two of your nine branches offer has already routed themselves.
The customer who messaged the wrong branch is still a customer, not a routing failure. Transferring them with "you have reached the wrong number, please contact our Deira branch on..." is the fastest way to lose them. The receiving branch should answer the question, take the booking, and hand the record over internally. Whoever is measuring branch performance has to be able to see a lead that arrived at one location and was served by another, or your team will start refusing to help.
"Any available agent" usually loses to "an agent from this branch." Round-robin assignment across all staff looks efficient on a dashboard. It fails the moment the question is local: which chair is free at four, whether the JLT parking barrier is working, whether the Marina branch has that colour in stock. A person answering for a location they are not standing in either guesses or goes to ask, and both take longer than the customer will wait. Route to the branch first, then to whoever inside that branch is free.
A closed or short-staffed branch needs a written rule, not a judgment call. Ours is simple and it has never been improved on: if the branch the customer wants cannot answer inside the promised window, the agent says so plainly, offers the nearest open location with a real slot, and offers the original branch's next available time as the alternative. Silence while someone "tries to reach the manager" is the only outcome that guarantees a loss. Set the fallback per branch, per day of the week, and let it fire automatically.
The second message should never re-route. When a customer comes back, the branch is already in the CRM record and the conversation should open where it left off. We run a 48-hour rule: inside 48 hours the agent treats the previous context as live and continues; past that it re-confirms the essentials, because prices, availability and intent all drift. Groups that route every inbound message from scratch make returning customers repeat themselves, which is the single most reliable way to make an automated system feel worse than a human.
The people cost of getting this wrong is measurable. On our own numbers, the recurring drain has two parts: roughly AED 2,000 a month of salary spent copying, chasing and re-explaining things a machine should handle, and around AED 4,100 a month of value lost to enquiries answered too slowly or not at all. Multiply the second figure by the number of branches before you decide routing is a nice-to-have.
Where the routing logic should actually live
There are three candidate homes and the wrong one is chosen constantly.
In the provider's inbox. Platforms like WATI, Respond.io, Trengo and SleekFlow all ship rules that assign conversations by keyword, team, shift or language. This is fine for two or three branches and one team structure. It breaks when routing depends on data the inbox does not hold, which is most of the time: stock, a doctor's calendar, whether this customer has an open complaint.
In a webhook layer you control. One WABA, several numbers, and your own service reads which number received the message and dispatches accordingly. This is the cleanest technical answer for Option B and the one BSP documentation describes. It also means you own a piece of software, which is a real commitment rather than a configuration.
In the CRM or booking system. The routing decision is made where the answer already lives: which branch has the slot, which branch owns this contact, which branch is open right now. This is the one we run, and it is the only version where routing and availability cannot disagree with each other. The practical requirement is that WhatsApp and the CRM have to be genuinely connected rather than exported nightly, which is a separate build in itself; the integration patterns are covered here.
One distinction worth naming, because these two problems get solved with the same software and are not the same problem. Several staff answering one number with ownership, assignment and a per-agent reply log is a shared inbox question. Sending an enquiry to the right location before anyone opens it is a routing question. A group with four branches and five people per branch has both, and solving the inbox alone leaves every conversation landing in one undifferentiated pile.
Meta's limits and the UAE paperwork
New business portfolios start capped at two registered phone numbers. Meta raises that to 20 automatically once your business is verified. Vendor documentation quotes slightly different ceilings above 20, so if you are planning for more than twenty branch numbers, get the current number from your provider in writing rather than from a blog.
Business verification in the UAE runs on a valid trade licence, the Emirates ID of the authorised signatory, and business contact details that match. Providers report one to five working days when the paperwork is consistent, and considerably longer when the name or address on the licence does not match what is in the Business Manager. Template approval is usually measured in hours rather than days once verification clears.
Each number carries its own display name, approved separately, which is what makes Option B viable: one account, twenty numbers, twenty names a customer can actually recognise. Display names can be changed a limited number of times in a rolling 30-day window and each change goes back through review, so get the naming convention right before you register nine of them.
Two UAE-specific points. Landlines can usually be registered through voice-call verification, which matters if your branch numbers are already printed on signage; confirm the specific number with your provider first. And consent for marketing templates has to be recorded per branch, because the branch that collected the opt-in is the one that can evidence it. Meta's own opt-in policy binds regardless of channel, and unsolicited commercial messaging in the UAE carries its own separate rules, so keep the record at location level rather than in one group-wide list.
What the whole thing costs
Platform costs are small and predictable: AED 40 to 75 per number per month for hosting, plus Meta's per-message rates, where marketing templates run roughly AED 0.13 to 0.20, utility and authentication roughly AED 0.03 to 0.06, and service messages free to 1,000 per number then billed at utility. A full cost breakdown sits in the API pricing guide.
The build is where the money goes. Wiring one process onto autopilot, which is what branch routing is, starts at AED 6,000 to set up and AED 1,200 a month to run on our automation service, and that connects what you already have rather than replacing it. A full agent that qualifies, prices, books and takes payment across branches is a larger build; what it includes is on the WhatsApp AI agent page. The same stack runs our studio business at about AED 67,600 net a month on roughly 5x ROAS, which is the only case study we can show you the back end of.
If you are not sure which of the four options your structure needs, that is exactly what a growth audit resolves in a week, from AED 3,000, with the number count and the routing map as the deliverable.
