WhatsApp calling has a reputation for being free. In personal use that is largely true: you pay no per-minute carrier fee. On the business side the picture is more complicated, because the cost of a voice conversation is not just the per-minute rate.
This post breaks down what makes up cost per call and compares the two channels in one table. The goal is not to give you a price — it is to give you the framework to produce your own number.
What a voice conversation actually costs
The cost of a voice conversation has four components:
- Transport cost. Getting audio from one place to another. Per-minute charges on GSM, data on WhatsApp.
- Handling cost. Who holds the conversation. An hourly rate for a human agent, a per-minute processing cost for a voice agent.
- Infrastructure cost. Lines, PBX, platform, integration, maintenance.
- Missed call cost. The business lost when a call goes unanswered. Never written down in most calculations, and usually the largest line.
Comparing only on the first component is the most common mistake. Zeroing out your per-minute rate while calls still go unanswered gains you nothing.
Channel comparison
| Item | GSM line | WhatsApp call |
|---|---|---|
| Per-minute charge | Carrier tariff | None, carried over data |
| International calls | High, varies by country | No additional charge |
| Line capacity | Limited by number of lines | Concurrency handled platform-side |
| Audio quality | Predictable, limited bandwidth | Clearer on a good connection, uneven on a bad one |
| Setup | Requires a line and a PBX | An existing WhatsApp number is enough |
| Caller identity | A number only | Number matches a customer record |
| Customer reach | Anyone can call | Customers who use WhatsApp |
| Outbound freedom | Broad, within legal limits | Within Meta policy and permission |
What the table shows is not that one replaces the other. A GSM line gives you reach and predictability; WhatsApp gives you cost and context advantages. The right setup connects both to the same agent.
The cost-per-call formula
To produce your own number:
Cost per call = (Monthly transport + monthly handling + monthly infrastructure) ÷ monthly calls answered
The critical word is in the denominator: answered. Not received. If 2,000 calls come in monthly and you answer 1,400, you divide by 1,400. The 600 unanswered calls do not lower your cost — they raise your cost per call, because you paid for the infrastructure and got nothing back.
Let us walk an example. The figures below are purely illustrative; replace them with your own data:
- Monthly inbound calls: 2,000
- Average call duration: 2 minutes
- Answer rate with a human team: 70 percent (1,400 calls)
- Answer rate with a voice agent: 100 percent (2,000 calls)
With a human team: the cost of a three-person team plus line costs, divided by 1,400 answered calls. With a voice agent: platform cost plus per-minute processing, divided by 2,000 answered calls.
Cost per call falls in the second setup for two reasons: the denominator grows, and team cost leaves the equation. More importantly, the business lost on those 600 unanswered calls enters the table.
Missed calls: the line everyone skips
The simple way to calculate the cost of a missed call:
Monthly loss = missed calls × average customer value × phone conversion rate
Do not estimate the conversion rate; pull it from your CRM. Someone who reaches you by phone is usually the person closest to buying, so that rate runs noticeably higher than for web form leads.
Once you add this line to the table, the comparison stops being a channel question for most businesses and becomes a capacity question. If the channel is cheap but calls still go unanswered, there is no gain.
Duration is a cost line too
Call duration is a direct cost under per-minute pricing. This is where the voice agent's architecture comes in: latency added on every turn lengthens the call.
Measured on our phone line, the model composes its answer in 0.65 seconds (659 and 624 milliseconds). In a chained architecture the same work is spread across three conversion steps, each adding its own share.
To see how the difference turns into cost, spread it across a call: on a ten-turn conversation, half a second per turn adds roughly 5 seconds. Negligible on one call; across 2,000 calls a month it is roughly three extra hours of talk time. Under per-minute pricing that goes straight onto the bill, and in customer experience it goes onto waiting.
When you decide
You can reduce the channel decision to three questions:
- How many of your customers are abroad? If it is a high share, WhatsApp calling closes a cost line on its own.
- What share of your calls goes unanswered? If it is high, solve answering capacity first; the channel is secondary.
- Where does post-call data go? Without recording, transcript and CRM write-back you are losing information no matter which channel you talk on.
You can see package and usage lines for your own calculation on the pricing page.