Skip to content

Bananaflow MCP is live.

Read the docs

Call Center Cost Per Call in India: Formula

Jaseem Abbas, founder of Bananaflow, in glasses and a dark top against a pale background.Jaseem Abbas

Published
Updated
Cover for "Call Center Cost Per Call in India: Formula": a white line diagram of one cost spread over different counts: a lit node on the left labelled Seat Rs 30,250 sends three arrows to three rows of dots, nine dots labelled Rs 8.40 a dial, three labelled Rs 28.01 a call and one labelled Rs 280 an outcome, drawn over a soft, out-of-focus petunia in muted sky blue, its broad ruffled petals opening sideways, with a cool mint haze at the left edge and a narrow buttercup glow.

A lead-calling floor in Pune dials 3,600 numbers a month from one seat, and 1,080 of them are answered. The seat costs Rs 30,250 a month. Is the cost per call Rs 8, Rs 28 or Rs 280? All three are correct, because each answers a different question. The number you need depends on whether you are judging dialling effort, conversations or results.

This guide shows how to work out the cost per call and the cost per minute for a call center in India, with one formula for each and one worked example. It also covers the five levers that move the number and which of them are worth pulling first. It builds on what a telecaller seat really costs compared with AI minutes, which explains where the Rs 30,250 comes from.

One seat, one month, three counts. The funnel is not drawn to scale; the counts and costs are from the worked example.

What is the cost per call in a call center?

Cost per call is the total cost of running the calling operation for a period, divided by the number of calls in that period. The catch is the word "call". A dial that rings out, a ten-minute conversation and a sale are all calls to someone, so state which one you are counting before you divide.

Most call centers report one number and use it for everything, which hides the real picture. A team that is great at reaching people and poor at closing has a low cost per dial and a very high cost per outcome. A team that talks slowly and thoroughly has a high cost per connected call and, sometimes, a lower cost per outcome. Judge each team on the measure that matches its job.

What is the cost per minute?

Cost per minute is the total cost of a seat or a team divided by the minutes spent in connected conversation, not by the minutes the person is logged in. A person who is paid for 8 hours and talks for 3 has 180 connected minutes a day, so every cost is spread over 180 minutes and not over 480.

This is the number to compare with a per-minute price from any vendor, including an AI voice agent. If you want the seat build-up, with salary, PF and ESI, the desk, supervision and attrition, it is in our guide to telecaller cost against AI calling cost.

Which formulas should you use?

Use five formulas: cost per minute, cost per connected call, cost per dial, cost per outcome and blended cost. Each divides the same total cost by a different count. Pick the one that matches the decision in front of you, and report the others beside it so nobody is misled.

  • Measure: Cost per minute

    Formula
    Total cost ÷ connected talk minutes
    What it answers
    What does an hour of real conversation cost?
    Watch for
    Count talk time, not login time
  • Measure: Cost per connected call

    Formula
    Cost per minute × average handling time in minutes
    What it answers
    What does one conversation cost?
    Watch for
    Short calls look cheap even when they fail
  • Measure: Cost per dial

    Formula
    Total cost ÷ dial attempts
    What it answers
    What does one attempt to reach a person cost?
    Watch for
    Falls when more numbers are answered
  • Measure: Cost per outcome

    Formula
    Total cost ÷ outcomes (appointments, payments, sales)
    What it answers
    What does a result cost?
    Watch for
    The only number tied to revenue
  • Measure: Blended cost per minute

    Formula
    (People cost + AI minute cost) ÷ all talk minutes
    What it answers
    What does a mixed team cost per minute?
    Watch for
    Needs the same cost lines on both sides

Average handling time, or AHT, is the average length of a conversation including any wrap-up work after it. Our AHT guide covers the formula, after-call work and benchmarks, and our call center metrics guide defines the other measures in the formulas.

What does the worked example show?

One seat costs Rs 30,250 a month. The person works 24 days and talks for 3 hours a day, which is 4,320 talk minutes. The average call runs 4 minutes, 30% of dials are answered, and 10% of answered calls reach the outcome. These are assumptions for the example, so replace them with your own report.

  • Count: Talk minutes

    How we get it
    24 days × 180 minutes
    Number a month
    4,320
    Cost per unit
    Rs 7.00 per minute
  • Count: Connected calls

    How we get it
    4,320 ÷ 4 minutes
    Number a month
    1,080
    Cost per unit
    Rs 28.01 per connected call
  • Count: Dial attempts

    How we get it
    1,080 ÷ 30% answered
    Number a month
    3,600
    Cost per unit
    Rs 8.40 per dial
  • Count: Outcomes

    How we get it
    1,080 × 10%
    Number a month
    108
    Cost per unit
    Rs 280.09 per outcome

The same Rs 30,250 produces four honest numbers. If you tell the owner "our cost per call is Rs 8.40", you are describing a dial that may have rung out. If you tell the sales head "Rs 280 to get an appointment", you are describing something they can compare with what an appointment is worth.

Now put AI minutes through the same counts. At 4,320 talk minutes and a 4-minute call, an AI agent handles the same 1,080 connected calls. At Bananaflow's pay-as-you-go rate of Rs 4.00 a minute plus 18% GST, which is Rs 4.72 a minute as of November 2026, one connected call costs Rs 18.88. On Growth it costs Rs 16.52, and on Scale Rs 14.16. At a 10% outcome rate, the pay-as-you-go cost per outcome is Rs 188.80.

Carrier charges are left out of every figure above, since a person and an AI agent both use the phone line. Add your carrier's per-minute rate to both sides. Prices are on the pricing page and may change.

What goes into the total cost?

The total is everything you spend to keep the seats working, not just salary. Six groups cover almost all of it, and the first three usually make up the bulk.

  • People. Salary and employer PF and ESI. The PLFS Annual Report for 2025 puts average earnings in regular salaried work at Rs 24,217 for men and Rs 18,353 for women across all sectors. It is a useful outer anchor, not a call center rate. On ESI, the ESIC contribution page states a 3.25% employer share.
  • The seat. Computer, headset, desk, power and internet, and software licences per person.
  • Supervision and quality. The team lead's and quality checker's time, spread over their agents.
  • Telecom. Carrier charges per minute, number rental and any trunk costs. For one carrier's rates, see Twilio pricing in India. Another carrier's plans are covered in Exotel pricing.
  • Technology. Dialer, CRM and reporting licences if they are not already in the seat cost.
  • Attrition and training. Recruiting, induction and the ramp-up weeks for new joiners.

Leave out costs that do not change with calling, such as the founder's salary. Include a share of any cost that exists only because the call center does.

Why do the three "per call" numbers differ so much?

They differ because each one divides by a different funnel stage, and each stage is far smaller than the one before. Dials are the widest stage, connected calls the middle, outcomes the narrowest. The same total cost spread over fewer units gives a larger number per unit.

This matters for a decision. If your connect rate is poor, because calls go to the wrong number or at the wrong hour, the cost per dial looks cheap while the cost per outcome is high. Fixing the number list and the calling hours then helps more than coaching your best agent. If your connect rate is high but outcomes are rare, the problem is in the conversation, and the script, the offer or the call timing is the place to look.

Which levers lower the cost per call?

Five levers move the number: talk time per day, average handling time, attrition, the connect rate and the share of calls that go to AI minutes. They differ a lot in size, so pull the large ones first.

  • Lever: Talk time per day

    Change made
    Raise from 3 hours to 3.5 hours
    Effect in the worked example
    Cost per minute falls from Rs 7.00 to Rs 6.00
  • Lever: Average handling time

    Change made
    Cut from 4 minutes to 3.5 minutes
    Effect in the worked example
    Cost per connected call falls from Rs 28.01 to Rs 24.51
  • Lever: Attrition

    Change made
    Cut from 5% to 3% a month
    Effect in the worked example
    Cost per minute falls from Rs 7.00 to Rs 6.96
  • Lever: Connect rate

    Change made
    Call in the hours people answer, and retry
    Effect in the worked example
    More talk time from the same hours, so the talk time lever applies
  • Lever: Move repeatable calls to AI

    Change made
    In a 10-seat team, move 40% of minutes to AI at Rs 4.72
    Effect in the worked example
    Blended cost per minute falls from Rs 7.00 to Rs 6.09

The moves are worth reading in order. Talk time is the biggest because the seat cost is fixed: every extra connected minute spreads it thinner. Attrition is the smallest in this example, though it also hurts quality and training time in ways the sum does not show. The AI lever changes the shape of the cost, replacing a fixed seat with a per-minute charge, and works by freeing whole seats.

For the AI row, the arithmetic is this. Ten seats cost Rs 3,02,500 a month for 43,200 talk minutes. Move 40% of those minutes, or 17,280, to AI at Rs 4.72, and you need six seats and AI: 6 × Rs 30,250 is Rs 1,81,500, plus 17,280 × Rs 4.72 which is Rs 81,562. The total is Rs 2,63,062 for 43,200 minutes, or Rs 6.09 a minute. It assumes you actually reduce the seats, and that the 40% is made of calls an agent can handle.

Each lever before and after its change in the worked example, each pair scaled to its own starting cost. The connect rate works through talk time.

How do you measure this on your own floor?

Start with the report you already have and fill the gaps with a short sample. A dialer or CRM report usually gives dials, connected calls and handling time. Talk time and outcomes may need a manual count for one week, then a regular report.

  1. Pull one full month of dials, connected calls, total talk minutes and outcomes from the dialer, the CRM or the telephony report.
  2. Add up the seat cost lines for the same month from payroll and the finance sheet.
  3. Divide as in the formula table, and write down all four numbers, not one.
  4. Pick the single lever with the biggest effect, change only that, and measure again next month.
  5. Repeat for each team or campaign, since a reminder campaign and a sales campaign should never share a cost per call.

Change one lever at a time. If you raise talk time, shorten the script and change the list in the same month, you will not know which one worked. For measuring talk time, handling time and connect rate the same way every month, see how call centers use AHT, FCR, CSAT and the rest. If the outcome you measure is an appointment, see how AI handles appointment confirmation calls.

How does the cost per call work for an AI voice agent?

For an AI voice agent, the cost per call is the call's length in minutes times the per-minute rate, plus GST and the carrier's charge. There is no fixed seat to spread, so it does not fall as talk time rises. It scales one for one with the minutes you use.

Because of that, the cost per outcome is where an AI agent needs the most attention. A call that runs long without reaching the outcome costs the same per minute as one that works. Test the script in the browser before it takes calls, and read the call records afterwards. Bananaflow keeps a transcript, a recording and the minutes used for every call, and it can score calls with a QA step, so you can see which calls ran long and why. See how analytics reports work.

How Bananaflow fits

Bananaflow lets you build AI voice agents that run campaigns from a CSV of numbers, with retries, calling time slots and a limit on calls at once. It charges per minute at the rates in the table above, and phone carrier charges are billed separately by your carrier. You bring your own numbers.

If you want to see how a campaign handles retries and call windows, read about campaigns and the outbound calls solution. This post is a method with sample numbers, so replace each one with your own before you decide anything.

FAQ

Frequently asked questions

  • How do you calculate cost per call in a call center?

    Divide the total cost of the seats for a period by the number of calls, and say which calls you count. Cost per minute is total cost divided by connected talk minutes. Cost per connected call is cost per minute times average handling time.

  • What counts as a call when you work out the cost?

    Pick one of three: a dial attempt, a connected call or a call that reaches the outcome you wanted. They give very different numbers. In our example the same Rs 30,250 seat costs Rs 8.40 per dial, Rs 28.01 per connected call and Rs 280 per outcome.

  • Which lever lowers cost per call the most?

    Talk time per day, in most cases, because a seat is a fixed cost. In our example, raising talk time from 3 to 3.5 hours cuts cost per minute from Rs 7.00 to Rs 6.00. Cutting attrition from 5% to 3% a month barely moves it.

  • How is the cost per call for an AI voice agent worked out?

    Call length in minutes times the per-minute rate, plus 18% GST and your carrier's charge. At Bananaflow's pay-as-you-go rate of Rs 4.72 a minute with GST, a 4-minute call costs Rs 18.88 before carrier charges. There is no fixed seat to spread.

  • What should be included in call center cost?

    People cost including employer PF and ESI, the seat and tools, supervision and quality, telecom charges, technology licences, and attrition and training. Leave out costs that do not change with calling.

Written by

  • Jaseem Abbas, founder of Bananaflow, in glasses and a dark top against a pale background.

    Jaseem Abbas

    Founder, Bananaflow

    Jaseem founded Bananaflow to build voice AI that understands how people in India really speak, including the switch between languages in the middle of a call.

Similar articles

Bring us one process. We'll pilot it on your real calls, with your team as the fallback.

Plan a pilot