All Products 
.svg){kind=icon}
Cost per call is total contact center operating cost divided by total calls handled over the same period. A center spending $140,000 in a month to handle 20,000 calls is running at $7.00 a call. Everything difficult about call center cost per call sits in the first term, because what you include in operating cost decides the answer, and most published figures never say what their authors included.
This page gives you the formula, a worked example built on federal wage data you can check, and an honest account of why the industry averages you have seen quoted are worth less than they look.
The Formula
Cost per call = total operating cost for the period ÷ total calls handled in that period
Keep the two terms on the same clock. Monthly cost over monthly volume, or annual over annual, never a monthly cost divided into an annual call count, which is a surprisingly common error in spreadsheets inherited from somebody who left.
What Belongs in Operating Cost
Three buckets, and the third is where most models go wrong by omission.
- Direct costs are the ones that obviously attach to handling interactions: agent wages, employer payroll costs, telephony and carrier charges, and licensing for the contact center platform, CRM and dialer
- Indirect costs keep the operation running without touching a call: supervisors, quality analysts, workforce planners, trainers, recruitment, facilities, hardware and IT support
- Variable costs move with volume or season: temporary staffing, overtime, additional language cover, and any compliance or legal overhead that scales with campaign activity
Leave out the indirect bucket and your cost per call will come in roughly a third low, which is exactly why one vendor’s published figure and another’s cannot be compared.
A Worked Example You Can Check
Start with labor, because it dominates. The Bureau of Labor Statistics puts the median hourly wage for customer service representatives at $21.53 as of May 2025, under occupation code 43-4051.
Wages are not the employer’s cost, though. In the Bureau’s Employer Costs for Employee Compensation release for June 2026, published on 9 September 2026, wages and salaries made up 68.4% of total compensation for civilian workers, with benefits accounting for the remaining 31.6%. Dividing the median wage by 0.684 gives a fully loaded labor cost of roughly $31.48 an hour before you have paid for a single license.
Now build the month, using twenty agents at 160 paid hours each.
| Line | Working | Monthly cost |
|---|---|---|
| Agent labor, fully loaded | 20 agents × 160 hours × $31.48 | $100,736 |
| Supervision, quality and training | Illustrative assumption for this example | $18,000 |
| Platform, CRM and telephony | Illustrative assumption for this example | $12,500 |
| Facilities, hardware and IT | Illustrative assumption for this example | $8,764 |
| Total operating cost | $140,000 | |
| Calls handled | 20,000 | |
| Cost per call | $140,000 ÷ 20,000 | $7.00 |
The labor line is built from published federal data and holds up to questioning. The three lines beneath it are placeholders for this illustration, not reference figures, and you should replace them with your own invoices and payroll before the number goes anywhere near a board pack. That distinction is the whole point: a model whose inputs you can name survives a finance review, and one assembled from figures in an article does not.
Why the Published Averages Fall Apart
You will find tables giving an average cost per call by industry: banking at one range, healthcare at another, retail lower still. They are tempting and they are close to useless, for three reasons.
They rarely state what was counted. A figure built on agent wages alone and one built on fully loaded operating cost can differ by 40% while both describe the same operation honestly.
They ignore geography. Labor is the dominant input, so the same call handled in a different market carries a materially different cost. A national average flattens that into nothing.
They blend call types. A two-minute delivery query and a twelve-minute fraud investigation requiring a senior agent, a compliance step and three systems are not the same product. Averaging them produces a number that describes neither.
There is also a sourcing problem worth naming. Most of the widely circulated per-call figures trace to commercial research sold by subscription or gated behind a form, so the methodology cannot be inspected, and the ones that are freely visible disagree with one another by multiples. Treat any industry average as rough orientation and never as a target.
Everything your team needs in one platform
Segment Before You Optimize
A single blended figure hides the only information you could act on. Three cuts do most of the work.
Split by purpose first. Inbound support, outbound sales, collections and retention have different handle times, different staffing patterns and different telephony profiles, and outbound in particular carries dialing attempts that never reach a person.
Split by channel second, because the per-interaction economics are not close. Voice is one-to-one, synchronous and the most expensive thing you run. Chat lets an agent hold several conversations. Email is asynchronous but often takes longer per case than people assume. Self-service through an IVR flow costs almost nothing per interaction once built. Reporting a blended cost across all of them tells you only that your channel mix changed.
Split by complexity third, using resolution effort rather than duration. A short call needing supervisor approval and three systems frequently costs more than a longer one resolved on a single screen.
Cost Per Resolution Is the More Useful Number
Here is the change that matters most to decisions. The purpose of a call is not to be short; it is to settle something. A fast call that leaves the issue open generates a second call, and the second call costs full price.
Cost per resolution divides the same operating cost by issues actually resolved rather than by calls answered. It punishes the false economy of pushing handle time down, and in sectors with long cases and specialist agents it is the only version of this metric that reflects what the money bought. Run both: cost per call for capacity planning, cost per resolution for deciding what to change.
Where the Money Actually Leaks
Three drains rarely appear as line items, which is precisely why they survive budget reviews.
Unproductive agent time. In outbound operations a share of paid time goes to calls that never reach a person, and the agent finds out only after listening to a greeting. Answering machine detection screens those before connection so agents spend their paid hours on conversations. The effect on your own cost per call depends on your connect rate, so measure it before and after rather than taking anyone’s multiplier on faith.
Fragmented workflows. An agent moving between a dialer, a CRM and a ticketing system pays a few seconds per switch, and those seconds multiply across thousands of interactions into real handle time. Embedding call controls inside the CRM the agent already has open removes the toggling rather than training people to do it faster.
Avoidable repeat contacts. When somebody rings back about the same issue, you pay twice for one resolution, and the second contact occupies a slot another customer needed. Repeat volume is not pure waste, since some callbacks are legitimate and expected, but avoidable repeats are the most expensive category of interaction in any operation. Speech analytics across the recorded set is how you find out which reasons keep recurring, rather than guessing from the ones that reached a complaint.
Reducing the Number Without Degrading Service
Four moves carry most of the benefit.
Deflect genuinely low-complexity contacts to self-service or messaging, where the marginal cost approaches zero, and leave voice for the interactions that need a person. Send a post-call SMS summarizing what happens next, since a caller who knows the answer does not ring back for it. Match staffing to the demand curve rather than to a flat rota, because overstaffing a quiet hour and understaffing a peak both cost money, in payroll and in abandoned calls respectively. Route by skill so specialists handle only what needs them.
What to avoid is equally clear. Cutting handle time without improving resolution moves cost from this week into next, where it reappears as repeat volume that nobody traces back. Watch cost per call, first contact resolution and customer satisfaction together on the same reporting view, because improving one of the three in isolation is almost always evidence that you have damaged another.
Frequently Asked Questions
How do you calculate cost per call in a call center?
Divide total operating cost for a period by total calls handled in that same period, so $140,000 against 20,000 calls gives $7.00. Include direct costs such as wages, payroll burden, telephony and licenses, indirect costs such as supervision, quality, training and facilities, and variable costs such as overtime and seasonal cover. Omitting the indirect bucket is the most common error and typically understates the result by around a third.
What is a good cost per call?
There is no portable answer, because the figure depends on your labor market, channel mix, call complexity and what you counted. Published industry averages rarely state their inclusions, ignore geography and blend call types, so they cannot be compared against your number meaningfully. A more useful target is your own trend, segmented by queue and call purpose, alongside cost per resolution so that falling cost is not simply moving spend into repeat contacts.
What is the difference between cost per call and cost per contact?
Cost per call counts voice interactions only. Cost per contact spreads the same operating cost across every channel, including chat, email and messaging, which usually produces a markedly lower figure because non-voice interactions carry a lower marginal cost. Neither is more correct, but mixing them is how reporting goes wrong: a center that migrates volume to chat will show a falling cost per contact and a flat cost per call, and both are telling the truth.
Does reducing average handle time reduce cost per call?
Mechanically yes, in the period you measure, and often not in the quarter. Shorter calls raise capacity per agent, which lowers the cost of each one. Where the time came out of resolution rather than out of inefficiency, the customer calls back, you pay again, and total cost rises while the headline figure improves. Read handle time beside first contact resolution, and treat a falling cost per call with a falling resolution rate as a warning.
How much of cost per call is labor?
Usually the clear majority, which is why the labor line deserves the most scrutiny. Build it from fully loaded cost rather than wages: federal data put the median hourly wage for customer service representatives at $21.53 in May 2025, while benefits accounted for 31.6% of total compensation for civilian workers in June 2026, giving roughly $31.48 an hour once both are counted. Using the bare wage understates every downstream figure in the model.
Should outsourced and in-house costs be compared directly?
Only with care, since the two are priced differently. An outsourcing rate per hour or per call usually bundles supervision, facilities, recruitment and technology that an in-house model lists separately, so comparing a vendor’s rate against your own agent wages flatters the vendor substantially. Build the in-house figure to fully loaded operating cost first, including the indirect bucket, then compare like with like and account for the volume commitments inside the contract.
Build the Model on Numbers You Can Defend
The hard part of cost per call was never the division. It is knowing what went into the top line, which queues are carrying the average, and whether a fall in the figure represents efficiency or deferred cost.
Talk to the Voiso team about the reporting behind cost per call and cost per resolution, and we will work through the segmentation against your own volumes rather than an industry average neither of us can verify.
