Skip to content

How to Build a Messenger Chatbot Flow That Sells

How to plan and build a Messenger chatbot flow that turns questions into orders: the building blocks, a six-step pet store example, drip sequences, testing, and the Meta rules every flow must follow.

Orange flow diagram of linked chat steps leading to a shopping bag, a Messenger chatbot flow that sells

Most chatbots fail when the customer asks a real question and the bot answers with a menu. A Messenger chatbot flow that sells works the other way round. It listens first, asks one question at a time, shows the right product, and hands over to a person when the sale needs a human touch. Here are the parts of a flow, a full pet store example, and the Meta rules that shape every message.

What a Messenger chatbot flow is made of

A flow is a set of steps that runs when something happens: a keyword, a button tap, a comment on a post. Each step does one job.

  • Message: text, an image or a document, such as a PDF size guide.
  • Buttons and quick replies: a few short choices the person can tap instead of typing.
  • Question: asks for a phone number or city and saves the answer.
  • Condition: picks a path based on an answer or a tag.
  • Wait: pauses the flow for minutes or hours before the next message.
  • Handoff: stops the bot and puts the conversation in front of a human.

Start with the sale, then build backwards

Before you open any builder, write down what a finished sale looks like in chat. For a cash-on-delivery shop: product and size chosen, name, phone and address collected, order confirmed. For a card shop: a click on a checkout link with the product already in the cart.

Then work backwards. Just before saying yes, people need the price, the delivery time and how to pay. Before that, they need the right product. That chain becomes your flow.

Example: a pet store flow in six steps

Picture a pet store in Lyon that sells food, beds and toys. Its most common message is "What food do you recommend for my dog?" Here is a flow that turns that question into an order.

  1. Greet and ask one question. "Happy to help. Is it for a puppy, an adult or a senior dog?" Three buttons.
  2. Ask about size. "Small, medium or large breed?" Each answer is saved as a tag.
  3. Show two products, not ten. A condition picks the right pair from the tags. Each card shows an image, the price and a "Choose this one" button.
  4. Remove the last doubt. "Delivery in 48 hours in Lyon, free from 49 EUR. Want a small trial bag first?"
  5. Close. Send the checkout link with the product added, or collect the address for local delivery.
  6. Follow up inside the window. If there is no click after two hours, a wait step sends one reminder: "Still deciding? I can compare the two foods for you." Then the flow stops.

At any step, a "Talk to someone" button sends the chat to the team. A dog with allergies needs a person, not a menu.

Write each message for a phone screen

People read Messenger on a phone, often with one hand. Keep each message under three short lines, and put the question last so the buttons sit right under it.

Use the words your customers use. If buyers in Abidjan write "c'est combien ?", that phrase belongs in your keywords and your reply should be in French. AI writing help can draft versions in several languages, but read each one before it goes live. A wrong price in translation costs more than the time saved.

Follow Meta's messaging rules

The 24-hour window. You can send any message, automated or not, within 24 hours of the person's last message. Keep every wait and reminder inside it.

After 24 hours. Only a human agent may reply, using the HUMAN_AGENT tag, for up to 7 days after the last message. It is for genuine support, such as a delayed parcel, never for a bot sending a discount.

Comments. A private reply to a comment is allowed once per comment, within 7 days.

Promotions. In Morocco, the EU and the UK, Marketing Messages are not available, so promotional broadcasts only reach people inside their 24-hour window. Recurring and one-time notifications have been retired. Details are in our article on the Messenger 24-hour rule.

Keep the conversation going with drip sequences

A drip sequence is a short series of messages sent over time. In Messenger it must fit inside the 24-hour window, so think in hours: an answer now, a tip an hour later, one last nudge before the window closes. Any reply from the customer resets the clock.

Use it to help, not to push. A tip on switching a dog to new food is useful. Three messages repeating the price are not.

Test, measure, then change one thing

Before you launch a Messenger chatbot flow, run it yourself from a personal account. Tap every button. Type letters in the phone field and check that the bot asks again politely.

After launch, look at where people stop. If many leave at the size question, it may be unclear or missing an option. Change one message at a time and compare week to week.

In IGNIPOST you build the flow from these same blocks: messages, images, documents, buttons, questions, waits, conditions and a handoff to your team. Handed-over chats arrive in the same inbox as your Instagram DMs and comments. Our Instagram DM automation playbook applies the same ideas to Instagram, and our unified social media inbox guide covers handovers.

This week, pick your most common customer question and build one flow for it: six steps at most, with a clear way to reach a person. When you are ready to set it up, compare the IGNIPOST plans or ask our team to look over your first draft.

Frequently asked questions

How many steps should a Messenger chatbot flow have?

For a first flow, aim for four to six steps from the first message to the order. Each step should do one job, such as asking a question or showing a product. Long flows lose people. If you need more information, collect it after the sale, or hand the conversation to a person who can ask for it naturally.

Can a chatbot message a customer after 24 hours?

No. Automated messages, including bot replies and reminders, are only allowed within 24 hours of the customer's last message. After that, a human agent can reply with the HUMAN_AGENT tag for up to 7 days, for genuine support questions. Plan every wait and follow-up so it lands before the window closes, or it will not send.

Should my chatbot try to answer everything?

No. A good flow handles the routine questions, such as price, sizes, delivery and payment, and passes the rest to a person. Complaints, custom orders and unusual requests need judgment. Put a "Talk to someone" button on every step, and make sure your team reads handed-over chats quickly during opening hours.

Share this guide

More in Social Media Automation, or browse all guides.

Put every channel in one calm workspace.

Schedule the week, answer every DM from one inbox, and let chat flows handle the questions you hear ten times a day.