Support metrics glossary
Cost per ticket. The number most teams compute wrong.
Cost per ticket is what one resolved support ticket actually costs you: the fully loaded cost of the support operation, divided by the tickets resolved. Most teams divide salaries by tickets and stop there. That is a labor rate. It is not a cost.
Last updated 13 September 2026
What cost per ticket means
Cost per ticket is the average amount a company spends to resolve one customer support ticket. The formula fits on one line: fully loaded support cost for a period, divided by the tickets resolved in that same period. Two words in that sentence do all the work. Fully loaded. Resolved. A ticket counts when the customer's problem is fixed, not when an agent clicks close. And fully loaded means every dollar the support function spends, not just the payroll line.
Cost per ticket calculator
One month, your numbers. The result is fully loaded; the salary-only figure sits next to it so you can see how far short the usual shortcut lands.
Monthly costs. Loaded means salary plus benefits and payroll taxes; overhead is team-lead, QA and training time.
Cost per ticket
$10.00
fully loaded, per resolved ticket
Fully loaded cost
$30,000/mo
Salary-only mistake
$6.67
($20,000 + $2,000 + $8,000) ÷ 3,000 tickets = $30,000 ÷ 3,000 = $10.00
What fully loaded actually includes
Three buckets. Agent compensation first: salary, benefits, payroll taxes. Then tools, the line everyone forgets: helpdesk seats, telephony, chat, AI licenses, the knowledge base. Then management overhead: team leads, QA, training, workforce planning. Skip the last two buckets and you get a number that feels precise and is not. Pricing support on salaries alone is like pricing a restaurant meal on the ingredients. It misses the kitchen, the rent and the person washing the plates, and it lands about a third under the real figure.
The math, worked through
One support team, one month. Four agents at $5,000 each, fully compensated, is $20,000. Tools come to $2,000. Management overhead, meaning a team lead, QA and training, adds $8,000. Fully loaded cost: $30,000. The team resolved 3,000 tickets. Cost per ticket: $10. The salary-only version of the same month reads $6.67, which is exactly the kind of number that gets a support budget cut for the wrong reason. Nothing in this table came from a benchmark report. Swap in your own lines and redo the division.
| Cost line | What goes in it | Amount |
|---|---|---|
| Agent compensation | 4 agents at $5,000 each, salary plus benefits and payroll taxes | $20,000 |
| Tools | Helpdesk seats, telephony, chat, AI licenses, knowledge base | $2,000 |
| Management overhead | Team lead time, QA, training, workforce planning | $8,000 |
| Fully loaded support cost | The three lines above, added | $30,000 |
| Tickets resolved | Problems fixed, not tickets clicked closed | 3,000 |
| Cost per ticket | $30,000 divided by 3,000 | $10.00 |
| The salary-only mistake | $20,000 divided by 3,000, with tools and overhead ignored | $6.67 |
The cost per ticket formula, and the industry average
cost per ticket = (agent compensation + tools + management overhead) ÷ tickets resolved
all four for the same period, and resolved means fixed, not closed
To calculate it, pull one month of payroll for the support team, salary plus benefits and payroll taxes, add the month's tool invoices and the share of team-lead, QA and training time that serves support, and divide by the tickets your helpdesk marks resolved with no reopen. Do it per channel if you can: an email ticket and a phone ticket cost different amounts, and the blended number hides which one is moving.
What is the industry average? We do not quote one here, because the published figures come from paid benchmark reports whose sample, cost definition and ticket definition you cannot see, and a service-desk password reset and an ecommerce return sit at different prices even in the same report. Benchmark internally instead. Take twelve months of your own figure as the baseline, split it by channel and ticket type, and judge every change against your own trend and your own resolution rate. A number that falls while resolution holds is a saving; a number that falls while reopens rise is a reporting artifact.
An AI agent changes one line of the formula: the tickets it resolves move out of the agent-compensation bucket and into tools, where Outlearn bills credits per AI reply, from 100 free credits a month to Starter at $99 for 2,000, so a resolved ticket costs cents of tooling instead of minutes of payroll.
The easiest metric in support to fake
Cost per ticket is the easiest number in support to lower, and the easiest to lower badly. Close tickets earlier and it drops. Make the contact form harder to find and it drops. Deflect customers into a bot that resolves nothing and volume falls, right up until those customers come back angrier through a channel you cannot deflect. Every one of those moves makes the metric better and the support worse. So the rule is simple. Never read it alone. Read it next to resolution rate and customer satisfaction, and treat any unconfirmed drop as a reporting artifact, not a win.
The levers that lower it for real
Start with agent time, because compensation is two thirds of the loaded cost. Anything that shortens average handle time lowers cost per ticket with it, which is why the two get reviewed together. Better docs, better tooling, cleaner handoffs. Then prevention, because the cheapest ticket is the one that never exists. In ecommerce, proactive shipping notifications remove the entire where-is-my-order category before it reaches the queue, and the WISMO page shows how large that category is. Finally, automation, but only the kind that resolves. A bot that answers the question removes cost. A bot that stalls just moves the ticket into next month's average.
Questions, answered first
What does cost per ticket mean in customer support?
The average fully loaded cost of resolving one support ticket. Add up everything the support operation costs in a period, salaries, tools and management overhead included, then divide by the tickets resolved in that period. Spend $30,000 in a month and resolve 3,000 tickets, and the cost per ticket is $10.
What counts as fully loaded support cost?
Three buckets. Agent compensation, meaning salary plus benefits. Tooling, meaning the helpdesk, telephony, chat and AI licenses. And management overhead, meaning team leads, QA, training and scheduling. Skip the last two and you are reporting a labor rate, not a cost.
Is there a benchmark I should compare against?
No honest universal one. The number swings with industry, channel and ticket complexity, because a password reset and a technical escalation can sit in the same queue at very different prices. Track your own figure monthly and judge changes against your own baseline.
How do you calculate cost per ticket?
Add the three cost buckets for one month: agents times their loaded monthly cost, tools, and management overhead. Divide by the tickets resolved in that same month. Four agents at $5,000 plus $2,000 of tools plus $8,000 of overhead is $30,000; over 3,000 resolved tickets that is $10.00 a ticket. The calculator on this page does the division with your numbers.
How is it different from cost per contact?
Cost per contact divides spend by every inbound interaction, including the ones that resolved nothing. Cost per ticket divides by resolved tickets. A customer who emails, chats and calls about one problem is three contacts and one ticket. Only the second number tells you what a fix costs.
Why use per-ticket pricing?
Because it turns the number on this page into a line on an invoice. Seat pricing charges for people whether or not they resolve anything, so cost per ticket has to be reconstructed after the fact. Per-ticket and per-resolution pricing tie the bill to the work done. Outlearn goes one step further and bills credits for the AI replies it actually writes, so a ticket that hands off to a person costs no resolution fee. The free plan carries 100 credits a month with no card, and Starter starts at $99 for 2,000 credits.
Why did our number drop after we added a chatbot?
Two possible stories. If the bot resolves issues, ticket volume falls and the saving is real. If it bounces customers back to the queue, the tickets return under new subject lines and the drop was cosmetic. Check reopen rates and repeat contacts before you believe it.