The short answer
A flat monthly plan costs the same in your busiest month as your quietest. Per-message and per-resolution plans cost most exactly when your marketing works or your knowledge base improves. At 500 conversations a month the two look similar; at 10,000 they are not close, and the gap is entirely in the vendor's favour.
A flat monthly plan costs the same in your busiest month as your quietest. A per-message or per-resolution plan costs more exactly when your marketing works. At 500 conversations a month the two models look similar. At 10,000 they are not close, and the gap is entirely in the vendor's favour.
The demo is always cheap. The invoice, six months later, is not. If you've shopped for a customer-support bot, you've met the pricing models: per message, per conversation, per "resolution," per seat. We sat down with real usage numbers from growing businesses and worked out what these models actually cost. The short version: the meter is the product.
What the same year costs on each model
Here is the arithmetic at three volumes. VaaniAPI figures are our own published flat pricing from $19/mo. The metered column is modelled, not quoted: WhatsApp Business API tools layer their subscription on top of Meta's per-conversation charge, and per-resolution tools bill each answered question separately.
| Conversations / month | VaaniAPI flat | Metered model | What changes at scale |
|---|---|---|---|
| 500 | $19 | Subscription + ~500 billed units | Roughly comparable |
| 2,000 | $49 | Subscription + ~2,000 billed units | Meter overtakes the flat plan |
| 10,000 | $119 | Subscription + ~10,000 billed units | Flat plan is unchanged; meter is 5x its own start |
We have deliberately not printed competitor figures. Per-conversation rates change by country, by conversation category, and by the month Meta revises them, so any number we quoted would be stale before you read it. Check the vendor's own pricing page on the day you buy, and compare what differs beyond price - grounding, citations and how handover works.
The three hidden multipliers
Per-message pricing looks harmless at low volume. But three things multiply it against you. First, growth: the moment your marketing works, your support volume - and your bill - doubles. Second, seasonality: a holiday sale can 10x your queries in a week, and the invoice follows. Third, hidden per-resolution overages that compound as your customer base expands.
Per-resolution pricing is subtler but worse: you pay more precisely when the bot works well. Your success has a meter on it.
If your support bill scales with your success, your vendor is your business partner - and you never agreed to that.
Why "per resolution" is the trap that reads best
Per-resolution pricing is the model that sounds fairest on a pricing page. You only pay when the bot actually solves something, so a bad month costs you nothing. Read it again from the other side: a good month costs you the most.
Work through what that does to your incentives. You spend money getting a customer to your site. They ask a question. The bot answers it well, and you are billed for that success. Every improvement to your knowledge base - every gap you close, every document you upload - increases the share of questions the bot resolves, and therefore your invoice. You are paying a fee that rises in proportion to how well the product works.
There is a second problem underneath it. "Resolution" is defined by the vendor, not by you. A conversation where the customer asked one thing and left may count. A conversation where they asked four follow-ups may count once, or four times, depending on whose billing logic you signed up to. That definition is worth reading closely before you commit.
The behaviour change nobody mentions
The most expensive consequence of metered billing is not the invoice. It is what businesses start doing to keep the invoice down.
Teams on metered plans told us they hid the chat widget on high-traffic pages, shortened the greeting so fewer people opened it, and in one case turned the bot off entirely during a sale - the exact week it would have earned its keep. That is a support tool being rationed. You bought something to answer customer questions and then arranged for fewer customers to ask.
Flat pricing removes the decision completely. Every extra conversation is free, so the only sensible strategy is to put the widget everywhere and let it work. That is a strategic difference, not an accounting one.
What predictable actually buys you
Predictable pricing isn't just cheaper on a spreadsheet - it changes behaviour. Businesses on metered plans told us they actively discouraged customers from chatting with the bot to control costs. Think about that: paying for a support tool, then rationing support. With a flat plan, every extra conversation is free upside. You want customers to ask questions, because answered questions become orders.
Seats, and the other meter
Volume is the meter everyone notices. Seats are the one that catches teams later. Per-seat pricing means the cost of letting a second person see the support inbox is another monthly line item, which quietly encourages exactly the wrong thing: one person holding all the context, and nobody else able to cover when they are away.
It matters more for a small team than a large one. A five-person company where only one person can open the inbox has a single point of failure, not a support process - one reason the enterprise helpdesks are a poor fit below a certain size. Worth checking what a plan includes before the second person needs access - ours lists seat limits per tier alongside the token allowance, and the escalation inbox is where those seats actually get used.
The festival-season problem
Indian retail does not have a flat demand curve. Diwali, Raksha Bandhan, end-of-season sales - query volume can go up several times over for a fortnight. On a flat plan that fortnight costs nothing extra. On a meter it arrives as an invoice in the month you were already spending on ads and inventory.
This is the point most pricing pages skip. The cost of metered support is not the average. It is the peak, and the peak lands at your worst possible cash-flow moment. Our festival-season playbook covers how to prepare the knowledge base for that surge.
How to compare tools honestly
Ignore the entry price. Model your cost at 3x your current volume - that's where you're headed if things go well. Ask whether the vendor wins when you grow, or only when you get billed. The right answer to "what will this cost me during my best month?" should be: exactly the same as your worst one.
Five questions worth asking every vendor before you sign:
- What is the total bill at 10x my current volume, including every pass-through charge?
- Is there a per-conversation or per-resolution fee on top of the subscription?
- Does a customer who asks three follow-up questions count as one billed unit or four?
- What happens in a month I exceed my plan - overage billing, or a conversation about upgrading?
- Can I cancel this month, or am I on an annual commitment?
What we charge, and what we don't
Our own answer: four plans, flat monthly pricing in USD, cancel any time. There is no per-message fee and no per-resolution fee, so a customer who asks eight questions costs the same as one who asks none. The full plan comparison lists token and seat limits per tier, and the billing FAQ covers what happens when you exceed one.
The honest caveat: token limits are a ceiling, and a very high-volume business will hit them. When that happens we contact you about upgrading rather than silently billing overage. That is still a limit - it is just a limit you can see coming.
Frequently asked questions
What is wrong with per-resolution pricing?
It bills you for the product working. Every gap you close in your knowledge base raises the share of questions the bot resolves, and therefore your invoice - and 'resolution' is defined by the vendor's billing logic, not by you.
Does VaaniAPI charge per message or per conversation?
No. Plans are flat monthly amounts in USD with no per-message and no per-resolution fee, so a customer who asks eight questions costs the same as one who asks none. Token ceilings apply per tier and we contact you about upgrading rather than billing overage silently.
Who charges the WhatsApp per-conversation fee?
Meta, on the WhatsApp Business API, regardless of which vendor's software sits on top. Hold that constant when you compare tools and compare the software price separately.
What should I ask a vendor before signing?
What the total bill is at ten times your current volume including pass-through charges; whether follow-up questions count as one billed unit or several; what happens in a month you exceed the plan; and whether you can cancel monthly.
