Your retailers are already on WhatsApp. The gap is between answering them by hand and running a system that takes orders while you sleep — and that gap is mostly setup decisions.
Most WhatsApp API projects go wrong at the first decision, because the question gets framed as "which vendor" instead of "what should this number do".
Write down the three or four jobs before you look at a single provider:
Every setup choice below follows from that list. If your list is only the last item, you may not need the API at all.
The free WhatsApp Business App is software you install. The WhatsApp Business API is an interface your systems talk to — there is no inbox, no notification, no phone. It only becomes usable once something is connected to it: a vendor's dashboard, a ticketing tool, or your own software.
That has two consequences people discover late:
Meta lets you connect through a Business Solution Provider or work with Cloud API directly.
| Through a BSP | Direct on Cloud API | |
|---|---|---|
| Time to first message | Days | Weeks, with developer time |
| Who handles Meta billing | The BSP, usually with a markup | You, billed by Meta |
| Template management | Dashboard | Your own build |
| Team inbox | Included | You build or buy it |
| Best for | Distributors without an in-house dev team | Companies with engineers who want full control |
For a distribution business the honest answer is usually a BSP to start. The markup buys you onboarding, a shared inbox and someone to call when template approval fails at 9 PM before a scheme launch. Move direct later if volume makes the margin worth the engineering.
This is the rule that shapes every workflow you will design.
When a retailer messages you, a 24-hour customer service window opens. Inside it, you can reply with free-form messages — text, images, PDFs, whatever the conversation needs. Outside it, you cannot start a conversation with free text at all. You must send a pre-approved template.
So the design question for every automation becomes: is this a reply, or an interruption? Replies are easy. Interruptions need a template, and templates need approval.
Templates fall into categories, and the category decides both approval strictness and cost:
The practical rule: if the message references a specific transaction the retailer initiated, it is utility. If it is trying to sell something, it is marketing, no matter how you word it. Miscategorising a marketing message as utility is the most common reason templates get rejected or accounts get quality-flagged.
Meta moved from conversation-based pricing to per-message pricing, and service conversations — the ones where you are simply replying inside the 24-hour window — are free. That pricing shift matters more than it sounds: it means answering retailers costs nothing, and only initiating costs money. Build your flows to earn replies.
A distributor running this properly has, in order:
New numbers start with a cap on how many unique users you can message in a rolling 24 hours, and the cap rises as you send without being blocked or reported. Quality rating drives the whole thing: enough retailers marking you as spam and your limit falls, or messaging gets restricted entirely.
The failure mode is nearly always the same. A distributor gets approved, imports 900 retailer numbers, and blasts a scheme announcement to all of them on day one. A chunk of those retailers never opted in, some tap "block", quality rating drops, and the number is limited before the flow has proven anything.
Do the opposite. Start with utility messages to retailers who already message you. Earn replies. Let the limit climb on its own.
Treat this as arithmetic to re-run with your own numbers, not as a benchmark.
Take a distributor serving 200 retailers, where a telecaller spends four hours a day taking and confirming orders by phone. At a ₹22,000 monthly salary across 26 working days and 8-hour days, four hours is roughly ₹1,270 a month of that person's time — call it ₹15,000 a year, plus the orders lost when the line is busy.
Now the other side: utility templates that confirm orders and invoices. If those same 200 retailers generate 2,000 order-related notifications a month, you are paying per initiated message for the ones that fall outside the 24-hour window — and nothing for the replies inside it.
The real return usually is not the telecaller's salary. It is the orders that arrive at 10 PM after the shop shuts, which no phone line was ever going to capture. Work out what one extra order per retailer per month is worth to you with the Profit Margin Calculator, and compare that against the messaging cost. If the number is not obviously positive, automate the reminders first and leave order-taking manual.
If you are starting from nothing, this sequence wastes the least time:
The temptation is to reverse this and launch with the AI ordering bot, because that is the exciting part. The distributors who do it that way tend to spend their first month debugging template rejections instead of taking orders.
Once messages are flowing, the next problem is designing flows retailers reply to — that is the subject of three WhatsApp workflows that drive revenue. And if your retailers type the way most Indian retailers type, half in English and half in Hindi, read why your commerce chatbot needs to understand Hinglish before you pick a bot.
Two separate things get billed, and conflating them is why messaging budgets come out wrong.
Meta's per-message charge applies to conversations you initiate — the templates above. Rates vary by country and by category, with marketing priced above utility. Replies inside a window the customer opened are service conversations, and those are free.
Your BSP's margin, if you use one. Some charge a platform subscription and pass Meta's rate through at cost; others mark up each message. Ask which model you are on before you sign, because at volume that difference is most of the bill.
The design consequence is the same one that keeps coming up: answering costs nothing, initiating costs money. A flow built so retailers message you first is structurally cheaper than one built on broadcasts, regardless of the rate card.
A worked example to re-run with your own numbers. Take 200 retailers. If each receives four utility notifications a month that fall outside an open window — order confirmed, invoice, dispatch, payment due — that is 800 initiated messages. Everything else, including the whole back-and-forth of taking the order, happens inside windows the retailer opened and costs nothing.
Now compare that against a weekly scheme broadcast to all 200: 800 marketing messages a month on its own, at the higher rate, to an audience that mostly did not ask. Same message count, several times the cost, and a quality-rating risk the first set does not carry.
Starting business verification last. It is the slowest step and everything waits on it. Begin the day you decide to do this, not the day you finish choosing a vendor.
Blasting the full retailer list on day one. New numbers carry a messaging limit that rises with clean sending history. Importing 900 contacts and broadcasting to all of them is the single most reliable way to get restricted before you have proven anything.
Submitting templates the morning of a launch. Approval is not instant and rejections are common on the first attempt. Get your core set approved in a quiet week.
No human fallback. Some conversations always need a person — a damaged consignment, a disputed balance. A system with no route to a human turns a small problem into an angry retailer.
Migrating your main number before testing. Moving a number to the API deletes its Business App history and you cannot go back easily. Prove the flow on a second number first.
The questions below come up in almost every setup conversation.
If one person can keep up with your message volume by hand, the free app is fine — it has broadcast lists, catalogues and quick replies. You need the API when you want messages triggered by your own systems (an order confirmation the moment stock is allocated, a reminder when an invoice ages past 30 days), when several people must answer from the same number, or when you want a bot to take orders. The dividing line is automation, not size.
The green tick marks an Official Business Account, which Meta grants based on how notable the brand is — press coverage and public presence. It is separate from Business Verification, which is the compliance check on your legal entity and is required to send at scale. You can run a full API setup, send lakhs of messages and never get a green tick. Most distributors do not need one.
Yes, but the number can only live in one place at a time. Migrating a number to the API deletes its chat history on the Business App, and you cannot use that number in the app afterwards. If the number is your main customer line, export what you need first. Many distributors start the API on a second number and migrate once the flow is proven.
Their reply opens a fresh 24-hour window, and you can answer freely again. The window only limits what you can initiate. Outside it you must use an approved template, which is exactly what template categories exist for.
The FlowKartAI team builds WhatsApp-native ordering for Indian B2B distributors and the kirana stores they serve. We write about distribution economics, GST compliance, and the practical side of putting AI in front of retailers who have never opened an app.
FlowKartAI parses natural language WhatsApp messages into ERP-ready orders in seconds.
Try Free Demo