What is cost per contact?

Cost per contact is the total cost of running a customer service operation divided by the number of customer contacts handled over the same period. It is the standard unit-economics measure for support, and it is usually expressed as a dollar figure per interaction.

Cost per contact is the number that turns operational choices into money. Staffing levels, channel mix, automation, self-service and training all show up in it eventually. That directness is what makes it useful, and also what makes it the most frequently misused metric in customer service.

This page covers what belongs in the calculation, how to work it out, what typical figures look like by channel, why the cheapest contact is not always the best one, and what to measure alongside it.

Cost per contact in one sentence

Cost per contact is what it costs, on average, to handle one customer interaction.

What goes into the calculation

The formula is simple. Agreeing what belongs in the numerator is not, and it is where most of the variation between organizations comes from.

Costs that are always included: salaries and wages for the support team, benefits and payroll taxes, and management and supervisor time.

Costs that are usually included: software licences for the platform and supporting tools, telephony and channel costs, training and onboarding, and quality assurance.

Costs that are sometimes included: recruitment, facilities and desk costs, IT support, and an allocated share of general overhead.

Costs that are usually excluded but arguably should not be: the cost of the engineering or operations work that fixes the root causes of contacts, and the revenue effect of the contact itself.

A cost per contact that includes only wages will be roughly half the figure of one that includes fully loaded overhead. Neither is wrong. Comparing one against the other is.

What counts as a contact

The denominator matters as much as the numerator.

A single customer issue that touches chat, then email, then a phone call may be one contact or three depending on how the platform counts. Systems built around tickets tend to count each new ticket. Systems built around conversations tend to count the whole thread as one.

This has a direct and slightly perverse consequence: an operation that counts every touch will report a lower cost per contact than one that counts whole conversations, from identical costs and identical customer experiences. It looks more efficient because it is fragmenting the denominator.

If cost per contact is being used for anything that matters, define a contact as a customer issue rather than as a system event, and hold the definition still.

How to calculate cost per contact

Cost per contact = total operating costs for the period ÷ total contacts handled in the period

A worked example. A support operation spends $100,000 in a month and handles 10,000 contacts.

Cost per contact = $100,000 ÷ 10,000 = $10.00

The same operation, if it counted each channel touch separately and recorded 14,000 touches instead, would report $7.14. Nothing changed except the counting.

Calculating by channel

The aggregate figure is the least useful version. Cost per contact by channel is where the decisions are.

Voice is consistently the most expensive assisted channel and self-service the cheapest, which is the observation the entire channel-shift argument rests on. Gartner put the median cost per contact at $13.50 for assisted channels and $1.84 for self-service in February 2024, and found that phone, chat and email cost broadly similar amounts as each other.

For voice specifically, ContactBabel's US Contact Center Decision-Makers' Guide has tracked the cost of an inbound call for more than a decade, surveying 189 US contact center managers and directors for its 2024 edition. It reported $6.91 in 2023, and across twelve years the figure never fell below $5.25.

Two caveats before benchmarking against either. ContactBabel notes that only a minority of its respondents answer the cost question at all, which suggests many organizations either do not measure it or are estimating. And the variation in what gets counted, described above, is wide enough that comparing your figure against a published average tells you less than comparing it against your own history.

Why the cheapest contact is not always the best one

Cost per contact has one structural flaw and it is a serious one: it counts contacts, not outcomes.

Consider two ways to handle the same 10,000 contacts.

In the first, the operation pushes hard on speed and channel shift. Cost per contact falls to $6. But 25% of those contacts were people getting in touch for the second time about an issue that was not resolved properly the first time. The cheap contacts generated the expensive ones.

In the second, conversations run longer and cost $11 each, but issues are resolved first time and the repeat rate is 8%. Total contact volume falls, because a resolved issue does not come back.

The second operation has a worse cost per contact and may well have a lower total cost. It certainly has better customers.

This is the trap. Cost per contact can be reduced by handling contacts worse, and the metric cannot see the difference. Every unresolved issue that generates a callback shows up as two cheap contacts rather than as one failure.

The metrics that close the gap:

Total cost of service, which is cost per contact multiplied by volume, catches the case where cheap contacts multiply.

Cost per resolved issue, which divides total cost by resolved issues rather than by contacts, is the version of the metric that cannot be gamed by fragmenting the work. It is harder to calculate and considerably more honest.

First contact resolution and contacts per order or per customer both show whether the volume itself is going in the right direction.

Read next to those, cost per contact is a good metric. Read alone, it will reward exactly the behavior nobody wants.

Cost per contact and the value side

There is a second thing cost per contact cannot see, and it matters more for some businesses than others.

A support conversation is not only a cost. For retail and ecommerce brands in particular, service interactions influence repeat purchase, order value and retention. A conversation where someone helps a customer find the right size, resolves a delivery concern, or saves an order is generating revenue while it costs money, and cost per contact records only the second half of that.

Optimizing a number that captures the cost of an interaction and none of its value produces a predictable outcome: the interactions get cheaper and the relationship gets thinner. Support organizations that are measured this way tend to discover the effect in the retention numbers a year or two later, where it is hard to attribute and harder to reverse.

The practical response is not to abandon the metric. It is to pair it with a value-side measure such as retention or repeat purchase rate for customers who contacted support, so the two halves of the picture sit on the same page.

How to reduce cost per contact without damaging the service

The interventions worth making are the ones that remove work rather than the ones that compress it.

Remove the contacts that should not exist. Repeat contacts about unresolved issues, and customers chasing information they should have been given proactively, are pure waste. Fixing the root cause removes cost permanently and improves the experience at the same time. This is the only lever that improves both sides of the ledger with no trade-off.

Let automation handle what it handles well. Order status, delivery updates, returns initiation, password resets and similar high-volume repeatable requests are resolved faster and more cheaply without a person, and customers generally prefer the instant answer. Genuine resolution matters more than routing here; a request that gets an automated non-answer and then a callback has cost more, not less.

Make each conversation more efficient without making it shorter. Time spent searching for order history, switching between systems, or re-establishing context the customer already gave is cost with no service value attached. Removing it lowers cost per contact and improves the conversation simultaneously.

Improve resolution rates. Every issue resolved first time removes a future contact from a future month's denominator. This is slower to show up than channel shift and it compounds.

Be careful with channel shift as a cost strategy. Moving volume to cheaper channels does lower the average. It lowers it most when the issue actually suits the cheaper channel and least when a complex problem is pushed into a channel that cannot resolve it, at which point the customer comes back through the expensive one anyway and both contacts are on the books.

Cost per contact when AI handles part of the volume

AI moves this metric more sharply than anything else in the last decade, and the reported numbers need reading with care.

The direct effect is straightforward. Contacts resolved automatically carry a marginal cost far below an assisted contact, so a blended cost per contact falls substantially as the automated share rises.

Three things complicate the picture:

The remaining human contacts get more expensive. Simple contacts leaving the human queue means the ones left are longer and more complex. Human cost per contact rises even when nothing about the team has changed. Reporting only the blended figure hides this, and it is the number that matters for staffing.

Resolution is what counts, not containment. A contact handled by AI without being resolved has not been removed from the operation, it has been deferred into next week at a higher total cost. The cost saving is only real where the issue actually went away. This is worth stating explicitly in any business case, because the two are easy to conflate and the difference shows up months later.

The value side does not disappear. The revenue influence of a good service conversation still exists, and an automated interaction that resolves the issue while losing the relationship has a cost that never appears in this metric.

The reporting discipline that keeps the number honest: report AI-handled and human-handled cost per contact separately, use resolved issues rather than contacts as the denominator wherever the data allows, and put total cost of service next to it so a falling unit cost on rising volume cannot be mistaken for a saving.