WhatsApp Business

WhatsApp passed 3 billion users in 2025, a number Meta confirmed on its 1 May earnings call. It is real, and it is why every vendor wants to sell you a WhatsApp programme. Read this chapter as the operator and compliance counterweight to those decks. The rule changes through 2025 and 2026 turned WhatsApp from a cheap broadcast channel into a metered, geo-branched, AI-gated one that punishes sloppy senders. Billing moved to per-delivered-template on 1 July 2025. Marketing templates to US +1 numbers have been paused since 1 April 2025; confirm no resumption before publishing. A per-user frequency cap means an opt-in no longer guarantees a delivery. For a Western business the default is not “WhatsApp instead of email”. It is WhatsApp as a high-intent, ad-led, conversational surface in specific markets, costed line by line before you commit.

This chapter leads with the mechanics that decide whether a programme works: the cost model, the per-user cap, the quality tiers, the template categories and the regional branching. There are no read-rate tables in here, and the first section explains why.

Kill the “98% open rate” before anything else

The “98% open rate” stat pinned to WhatsApp in every pitch is not a measurable number, and you should cut it wherever you find it. WhatsApp does not report opens. No open pixel, no open event in the API, nothing that says a message was seen.

The closest signal is the blue double-tick read receipt, and there are three problems with treating that as an open rate. The recipient can switch read receipts off entirely, and plenty do. When they are on, a business sees them per conversation, one thread at a time, never as an aggregate you can roll up across a send. And a read receipt is not an open in the email sense. It is a delivery-to-screen acknowledgement that the message rendered in a chat the user may never have looked at. Any “98% open” figure is a vendor extrapolation dressed up as a platform metric.

What WhatsApp will tell you is sent, delivered, and where read receipts are enabled, read. On top of that you have your own click and conversion tracking on the links inside your templates, the same as any other channel. Anchor your reporting and your expectations on delivered-and-billed plus your own downstream clicks and conversions. If a tool or a teammate quotes you a 98% open rate, they have not understood the channel.

The cost model, which is the part most people get wrong

Since 1 July 2025 Meta bills per delivered template message, not per 24-hour conversation. The old conversation-based pricing and the “1,000 free conversations per month” allowance are both gone. Any spreadsheet, vendor deck or blog post that still models WhatsApp on free conversations is wrong, and carrying it into a 2026 plan will throw your cost estimate off by a wide margin.

The rate for any given template is set by three things acting together: its category (Marketing, Utility or Authentication), the recipient’s country calling code, and, for Utility and Authentication, a volume tier that lowers the unit rate as your monthly volume climbs. Marketing gets no volume discount, which bites when you model a large promotional send. So the cost formula is:

total cost = sum over recipients of (delivered billable template messages × the applicable category / country / volume-tier rate) + BSP and pass-through fees

The “delivered billable” qualifier is doing real work. Three situations make a delivered template free rather than billable, and those three free lanes are the entire cost strategy on this channel.

The three free lanes

Trigger	Window	What is free	Operational move
User messages you first	24-hour customer-service window	Service messages and any free-form reply you send	Run support inside this window; it costs nothing
User messages you first (Utility category)	Same 24-hour open window	Utility templates (order confirmations, shipping updates, appointment reminders) delivered inside the open window; charged only when sent outside one	Send transactional WhatsApp while the user already has a live thread with you
User starts the chat from a Click-to-WhatsApp ad or a Facebook/Instagram Page call-to-action, and you reply within 24 hours	72-hour Free Entry Point window	All categories, including Marketing	Pair a Click-to-WhatsApp ad with an instant automated reply to open this window; it is the biggest lever on WhatsApp spend that exists

Meta has changed the Free Entry Point mechanics more than once, so re-confirm the exact duration and eligibility against the Meta pricing documentation on the day you lock your plan.

Everything outside those lanes is billable. Marketing templates sent cold, with no open window, are always billed at the marketing rate for the recipient’s country. Authentication and Utility sent outside a window are billable too, but volume-tiered, so they get cheaper as you scale. Rates are country-specific and they change. Live-fetch them from your BSP or the Meta rate card every time you build a model. Never hardcode a per-country rate; a number you wrote down last quarter is a liability, not a saving.

A worked example shows why the formula matters more than any single rate. Say you want to reach 50,000 opted-in contacts split across three markets with a promotional template. Under the dead conversation model your instinct would be “first 1,000 free, then a flat per-conversation fee”, and you would budget one number. Under the live model you split the list by country code first, apply that country’s marketing rate to each slice, accept that none of it qualifies for the volume discount because it is Marketing, then subtract every contact already over their per-user cap that day, who will bounce with 131049 before you are billed. The same 50,000-contact send can cost two or three times a naive flat estimate in expensive markets, and a fraction of it in a cheap-SMS market. There is no single “WhatsApp cost”, only a per-slice cost you assemble from the rate card on the day. Build the model as a per-country breakdown from the start and the invoice will not surprise you.

Cost reality: where WhatsApp pays off in the West, and where it does not

The vendor decks never say this plainly, so I will. WhatsApp marketing is billed per delivered message at a country rate, and in the West that rate sits well above SMS in the expensive markets. Germany and France marketing templates run several times the cost of a cheap US SMS. The UK sits roughly level to somewhat higher. Those are directional shapes, not figures to quote, and the only honest move is to pull the live rate card for the countries you actually send to before you decide anything.

Two structural facts make cold WhatsApp marketing a weak default for a Western business specifically. First, marketing templates to US +1 numbers have been paused since 1 April 2025; confirm no resumption before publishing. If your audience is American, WhatsApp marketing is not a lane to rely on. Second, the per-user cap (next section) means you cannot reliably reach an opted-in list with promotional templates even in the markets where marketing is allowed. So for broadcast promotions to a Western audience, email and SMS usually win on cost per message, and WhatsApp often loses the comparison before you have sent a single template. Do not stand up a costly WhatsApp blast programme expecting an SMS substitute. You will not get one.

Where WhatsApp earns its cost is the opposite of broadcast. The free lanes do the work. Replies inside the 24-hour service window are free, and a Click-to-WhatsApp ad answered within 24 hours opens a 72-hour window in which even Marketing is free. Pair a Click-to-WhatsApp ad with an instant auto-reply, and a high-intent conversation costs you the ad click, not a per-message marketing fee. That, plus the markets where WhatsApp simply is the default inbox (India, Brazil, the rest of LATAM, MENA, and parts of the EU), is the genuine case for the channel. A WhatsApp programme that pays for itself is conversational and ad-led. Treat per-message marketing templates as the expensive lane and the windows as the cheap one, and build accordingly.

Per-user caps: opted-in does not mean delivered

This is the fact that breaks most people’s mental model, so sit with it. Meta applies a per-user cap on marketing template messages. It is dynamic, engagement-gated, and counts across every business messaging that user, not just yours. Meta publishes no fixed numeric cap; the actionable signal is error 131049, which tells you a given send was refused. Some BSP reporting (Vonage among them) describes the practical ceiling as around two marketing templates per user per day summed across all brands, but treat that as anecdotal reporting rather than a published rule.

When a user is over the ceiling, your send fails with error 131049, the “healthy ecosystem” error, regardless of how good your own quality rating is. Read 131049 correctly: it means “too much low-value marketing is already reaching this person”, not “transient glitch, retry in five minutes”. Retrying it harder makes things worse. It is the clearest single proof that “opted in” and “delivered” have come apart on this channel. Utility and Authentication templates are not subject to the marketing cap, so your transactional and one-time-passcode traffic is unaffected.

The operator lesson is blunt. On WhatsApp, “opted in” and “delivered” are now two different states. A subscriber can have given you a clean, documented opt-in, and your marketing template to them can still be refused at the door because three other brands got there first that day. You cannot buy your way past this, you cannot warm your way past it, and it stays invisible until you read your error codes. Build your reporting so 131049 surfaces as its own line, not buried in a generic “failed” bucket, because it is telling you something specific about how crowded that inbox already is.

There are practical moves that work with the cap rather than against it. Move anything that can legitimately be Utility (order updates, reminders, account notices) into the Utility category, because it is exempt from the marketing cap and far more likely to land. Concentrate your one genuinely marketing send of the day on the segment most likely to act, rather than blasting the whole list and watching half of it bounce. And lean on the windows: a message sent inside an open service window or a Free Entry Point window is not a cold marketing template, so the cap is not a gate it has to clear. The brands that get good delivery on WhatsApp are not the ones sending the most marketing templates. They are the ones who have arranged their traffic so most of it never has to compete as a cold marketing template at all.

Quality-gated sending tiers

WhatsApp meters how many unique recipients you can reach in a rolling 24 hours through quality-gated tiers, and Meta is actively restructuring them, so treat any specific ladder as perishable. The historic ladder ran 250 unique recipients (unverified) up through 1K, 10K, 100K, to unlimited, climbed by sending to a healthy share of your current cap while holding a good quality rating.

Through 2026 Meta is collapsing the middle of that ladder. BSP reporting (Chatarmin among others) describes the 2K and 10K steps being removed, so a verified sender jumps toward 100K and then unlimited, with tier checks running roughly every six hours rather than once a day. Since around October 2025 the limit pools across a whole Business Portfolio rather than per phone number, and a new number inherits the portfolio’s highest tier, which is the structural reason to keep your numbers organised under one Business Portfolio. Re-verify the exact current ladder against Meta’s messaging-limits documentation when you lock your plan, because the version you read here will have moved.

Map all of this onto the deliverability chapter you have already read, because the mechanics are the same shape under different names. Quality rating behaves like sending-domain reputation. Blocks and reports are the WhatsApp equivalent of spam complaints. Tier climbing is warm-up. One useful nuance: a red quality rating no longer auto-downgrades your tier if you have no policy violations, but it does freeze your advancement, so you stop climbing until you clean up. The discipline that earns a good email reputation, relevance, restraint, sending to people who actually want it, is the same discipline that climbs WhatsApp tiers.

Template categories and auto-recategorisation

There are three billable categories, Marketing, Utility and Authentication, plus a free Service type for replies inside an open window. The category is not a label you assign and forget. Meta auto-recategorises templates it judges to be mis-filed, and bills accordingly. Any promotional language in a template you submitted as “Utility” will flip it to Marketing, at which point it is billed at the marketing rate and becomes subject to the per-user marketing cap.

Two practical rules follow. Ship promotional content as Marketing from the start, rather than dressing a promo up as a utility message and getting recategorised anyway. And stage your templates early, well before a campaign date, so that if Meta recategorises one you have an appeal window before the send rather than discovering it the morning of. Meta has tightened the advance-notice rules for repeat offenders, so re-verify the current recategorisation notice period against Meta’s documentation rather than relying on what a BSP restated last year.

Opt-in and regional compliance

An opt-in is mandatory before any proactive message. Under Meta’s current policy you may collect it on any channel, but it must name the business, state clearly what the user is opting into, comply with local law, and support opt-out both on and off WhatsApp. Click-to-WhatsApp ads and QR codes give you the highest-quality capture, because the intent is unambiguous and the consent context is clean. A user who taps a Click-to-WhatsApp ad and opens a chat has done something more deliberate than ticking a box at checkout, and that intent shows up later in lower block rates and a healthier quality rating. Where you can, prefer capture methods where the user starts the conversation, because those are the same methods that open your free windows. Treat your opt-in records the way the compliance chapter tells you to treat email consent records: timestamped, attributed to a source, and retained, so that if a regulator or Meta asks you can show when and how each person agreed.

Compliance then branches hard by region, and region should be a first-class input to every campaign, decided before you write a template, not bolted on at send time:

• US (+1): Utility, Authentication and service-window replies only. Marketing templates to US +1 numbers have been paused since 1 April 2025; confirm no resumption before publishing. Pivot US promotional traffic to email or SMS by default.
• India: Meta opt-in plus pre-approved templates, plus DPDPA obligations. India is not under the SMS-specific TRAI/DLT regime for WhatsApp, so do not drag SMS-DLT registration myths across onto this channel.
• EU: active, granular GDPR consent, a data-processing agreement in place, and ideally an EU-based BSP for data-residency comfort.
• Brazil: a documented LGPD lawful basis for the contact and the messaging.

These per-country specifics come from trade and BSP guidance, not a single regulator page, and the rules move. Treat the list as a starting checklist to confirm with counsel in each jurisdiction you send to. It is not legal advice, and not a substitute for advice from someone who carries the liability.

The third-party AI chatbot rule

Meta changed the WhatsApp Business API terms to bar general-purpose third-party AI chatbots, the ChatGPT, Perplexity and Grok class of assistant, from the platform, with enforcement set for 15 January 2026. This matters for an AI-native programme, so be precise about what it does and does not cover.

It does not stop a business from running its own scoped, task-specific agent. A customer-service bot, an order-tracking agent, a booking assistant, a lead-qualification flow, all of these are fine, because they are tied to your actual workflow rather than being a general-purpose assistant bolted onto the API. The rule is aimed at companies using WhatsApp as a distribution surface for a standalone AI product, not at businesses automating their own support.

The policy is also under live antitrust challenge and has already been carved out by country. The global rule is the bar on general-purpose third-party assistants. The carve-outs sit on top of it and differ by region. The requirement does not apply when messaging Brazil (+55) numbers, following a similar earlier exemption for Italy. The EU position is separate and moving: regulators there have reportedly intervened on competition grounds, with reporting that Meta has been ordered to restore rival-assistant access while the investigation runs, so the practical status in the EU is unsettled. Coverage is region-specific and genuinely in flux. Before you build anything that puts an LLM on the Business API, re-verify the current status of the ban, the Brazil and Italy carve-outs, and the EU position at publish time rather than assuming any fixed outcome. The build rule itself is simple and stable even as the edges shift: deploy a scoped business agent tied to a real workflow, and never bolt a general-purpose assistant onto the Business API.

Vendor and practitioner numbers, as sanity-check ranges

These figures are for sense-checking your own results, not for setting targets, and they should never appear as a benchmark table. They come from vendor studies and DACH e-commerce practitioners, and they describe a specific market that may not be yours.

The Chatarmin KPI study, drawn from 30-plus DACH e-commerce brands, reports broadcast read rates in the 60 to 80% band, click rates of 5 to 12%, conversion of 3 to 7%, opt-out around 0.1 to 0.2%, and broadly 15 to 40x broadcast ROAS, with automated flows running higher and the top of the ROAS range being outliers rather than the norm. Use that band to ask whether your own WhatsApp data is plausibly in range, and nothing more.

Click-to-WhatsApp ads are Meta’s fastest-growing ad format, and Meta reports strong year-on-year revenue growth for them, which is its own internal figure. The often-quoted “94% conversion lift” and “92% lower cost-per-lead” both come from a Meta-commissioned Forrester study, so treat them as directional and plan to your own market-dependent cost-per-conversation rather than the headline percentages.

On cadence, practitioner consensus lands at roughly two to three promotional messages per week, timed to local morning and evening and spread out rather than clustered. Meta publishes no cadence number of its own. The discipline matters more here than on email, because WhatsApp shares an inbox with friends and family. Over-message and you trigger mutes and blocks, which feed straight back into the quality-rating throttle and the per-user cap. The cost of annoying someone on WhatsApp is not just an unsubscribe. It is a measurable hit to your ability to reach everyone else.

Commerce capability menu

WhatsApp supports a real set of in-thread commerce features, and these are the genuine alternatives to a text blast: catalogs, single and multi-product messages, media-card carousels of up to 10 cards, Limited-Time-Offer templates, WhatsApp Flows for in-chat booking and checkout, and a Business Calling API added in 2025.

Match the feature to the job. Carousels for showcasing a range, Limited-Time-Offer templates for genuine urgency, Flows for cart and lead capture inside the chat without bouncing the user to a browser, and the Calling API for high-value handoffs where a conversation closes the deal. Used well, these turn a WhatsApp thread into a place where the transaction actually happens rather than a place you point people away from.

The named brand case studies you will see attached to these features (JioMart, SNOCKS, Skullcandy and the like) and the round-number engagement stats (“175M people message a business daily”, “40M catalog views per month”) are vendor-sourced. Keep them illustrative and clearly flagged as such. They are useful for showing a stakeholder what good looks like, not for setting your own targets.

Re-confirm the volatile mechanics at publish time

Everything in this chapter is current as of mid-2026 and verified against Meta’s developer documentation where a primary page exists. Meta revises these mechanics often, and a BSP restating an old rule is not a substitute for the source. Before you build or publish anything load-bearing, re-check the following against Meta’s own docs, not a third-party summary: the template recategorisation notice rules, the per-user marketing cap, the messaging-limit tier ladder (which Meta is collapsing through 2026), the third-party AI chatbot policy and its country carve-outs, and the Free Entry Point window duration and eligibility. And live-fetch every per-country rate at the moment you cost a campaign. The one habit that keeps your WhatsApp numbers honest is refusing to hardcode a single rate, ever.