How to adapt to WhatsApp’s per-message pricing with less friction for customers

Since Meta’s latest update, replying to a customer on WhatsApp costs money. Meta gives each number a free monthly allowance, and beyond that it charges for every message the business sends, whether a person or a bot writes it.

For high-traffic companies, especially in Latin America, the channel’s bill can double or triple. That’s why many are asking whether the channel “is no longer sustainable or no longer delivers the return on investment it used to,” in the words of Marc Caballé, Hubtype’s CEO.

Even so, it’s where a large share of their customers already are, and part of the bill depends on something the company does control: how its conversations are designed.

Why can a “hi” cost three messages?

“The reality is that companies used this channel because it didn’t cost anything,” Marc admits.

While that was true, few companies counted how many messages it took to serve a customer.

In fact, most conversations were designed to feel like a chat between two people, with short lines arriving one after another, and agents wrote the same way: first they let the customer know they’d seen the message, then they answered.

Now each of those messages adds to the bill, even if, for the customer, they make up a single reply. Country rates, message categories and the details of the free allowance are covered in our article on WhatsApp Business API pricing.

“If you say ‘hi’ and the bot replies in three separate messages, you pay three times,” Marc says.

The same goes for agents: if a customer asks where their order is and gets “hi, I’ll be right with you” followed by “your order is on its way,” that’s two messages. In both cases, the customer only needed one.

How do you redesign a conversation to pay for fewer messages?

Short answer: by cutting the messages that don’t bring the customer any closer to resolving their request.

Eric Marcos, Hubtype’s CTO, sets out where the line is: “reduce the number of messages the business sends as much as possible, while affecting the user experience and problem resolution as little as possible, which is ultimately the goal every business has.”

There are three points in a conversation where messages tend to pile up:

The welcome. Many assistants introduce themselves in three steps: “Hi, I’m Company X’s virtual assistant,” then the legal text, and finally the greeting. Each of those messages is a charge. If the welcome is designed as a single message, the customer reads the same thing in one go and the business pays for one.

“Always trying to combine these messages is one way to cut down,” Eric says.

The “we’re looking into it” message. “Ok, I’m looking into it” tells the customer someone is on their case, but it doesn’t solve anything, and now it costs money. If the answer comes quickly, the conversation can go straight to it. If the request will take a while, a single message with an estimated time covers what several updates used to. Keep in mind that emoji reactions are charged too, so they don’t save messages either.

The close. When a case is resolved, many companies send a satisfaction survey. “The vast majority go unanswered. So it’s a bit of a wasted message", Eric explains. Instead of the survey, the conversation can be analyzed with AI. “What we sometimes do is remove it and simply analyze the conversation with artificial intelligence. We now have a product that uses AI to evaluate the outcome of the conversation and the customer’s mood, and these end-of-conversation surveys often stop making sense,” Eric says. That way, satisfaction is measured across every conversation, not just the few where someone answers the survey.

These three changes cut individual messages. In some requests, though, you can do without far more than one message. When a request needs several details in a row, such as a change of address, a new insurance claim or a flight change, what you can do without is the chat format itself.

Which requests are worth taking out of the chat?

Short answer: the ones that need several details in a row, such as a change of address, a new insurance claim or a flight change. Over chat, they’re handled one question at a time: the bot asks for a detail, the customer types it in, and the bot asks for the next one.

Each question is a paid message and a moment when the customer might give up halfway through.

Those processes can move to a screen that opens inside WhatsApp itself, similar to a form. That’s what webviews and WhatsApp Flows make possible. Eric describes it as “bringing up a graphical interface on top of the chat,” where the customer can fill in a form or complete any other interaction “without using up paid WhatsApp messages.” And it works even for “a fairly long and fairly complex process, saving you a lot of messages.”

The customer notices the difference too. They see all the fields at once, can correct things before submitting and don’t have to wait for the next question, so they’re more likely to get to the end. The savings are most noticeable in conversations where a process like this accounts for a large share of the messages.

When does it make sense to route traffic to other channels?

Short answer: when those channels have significant adoption in your market.

Besides cutting messages within WhatsApp, part of the traffic can be handled on other chat channels that don’t have this per-message pricing, such as webchat or Facebook Messenger. According to Eric, they’re real-time channels that are “just as valid” for customer service.

That said, whether it pays off depends on how much your audience uses them.

“You need to factor in how many people use these channels in your market, but if adoption is significant, it makes sense,” Eric explains.

If your customers barely use another channel, their conversations will keep coming in through WhatsApp.

Up to this point, every measure is in the company’s hands. Meta, for its part, offers another option: its own AI agent.

Is Meta Business Agent an option for a large company?

Short answer: it can be for a small business, but it falls short for a large company.

Meta Business Agent is the AI agent Meta has built into WhatsApp. It lets businesses connect knowledge bases and serve customers with generative AI inside the channel itself. It has been billed per token since August 1.

For Eric, the timing of the price change isn’t a coincidence: “Meta is trying to push adoption of this new service a bit, by making the option without it more expensive.”

The problem lies in what a large company usually needs: meeting regulatory requirements, connecting sensitive data from its own systems and adapting the service to its processes.

It also needs to know which AI model is running underneath, and in Meta’s case, according to Eric, that’s unclear. That’s why he thinks the service “falls a bit short” for this kind of company.

If Meta’s agent isn’t a fit, there’s another conversation to have with Meta: the one about price.

Can you ask Meta for some leeway to adapt?

Short answer: in some cases, yes. It depends mainly on the relationship the company has with Meta.

That leeway buys time to put the measures above in place while conversations are being redesigned.

“They’re still in a testing phase, after all. We don’t know what’s going to happen, whether they’ll keep prices as they are or lower them, but it’s important that they hear the feedback,” the Hubtype team explains.

Companies that invest with Meta in advertising or other services can bring their case to Meta and ask for a grace period or a discount while they adapt.

Those without that relationship can raise it through their provider. As an official Meta Business Solution Provider (BSP), Hubtype has been passing its customers’ feedback on to Meta from day one and can help check whether a specific account qualifies for an exception.

Any leeway, if it comes, is Meta’s call. What depends only on the company is how it designs its conversations: cutting the messages that don’t help the customer lowers the bill, and a customer with fewer steps leaves fewer requests unfinished.