Cost per ticket is the total monthly operating expense of your support organization divided by the number of tickets it resolved that month. Published 2026 benchmarking puts the average IT service desk somewhere between $15 and $25 per ticket, with real organizations ranging from under $3 to nearly $50. Customer support teams usually price the same math per contact instead: roughly $9 to $16 for a phone contact, $5 to $9 for chat, $6 to $11 for email, and well under a dollar for a self-service resolution.
Almost every support team can tell you its handle time, its CSAT and its backlog. Far fewer can tell finance what one ticket costs. That gap matters, because cost per ticket is the number that converts every other support metric into money. Deflection is interesting; deflection multiplied by cost per ticket is a budget line. Handle time is interesting; handle time multiplied by loaded agent cost is a business case for better tooling.
It is also the number you need before anyone can honestly evaluate an outsourcing quote, an AI agent, a new ticketing platform, or a headcount request. Without it, every one of those conversations is a comparison against a figure nobody has calculated.
What is cost per ticket?
Cost per ticket is a unit economics metric: the fully loaded cost of running support, divided by the volume of work support completed. MetricNet, whose benchmarking is the source most service desk figures trace back to, defines it as the total monthly operating expense of a service desk divided by monthly ticket volume.
The word doing the work in that definition is total. Cost per ticket is not agent salary divided by tickets. If you only count the people who answer, you will land 30 to 50 percent below your real number, present it to finance, and lose credibility the first time somebody adds up the invoices.
The cost per ticket formula
The formula itself is a single line:
Cost per ticket = total monthly support operating expense / total tickets resolved that month
Everything hard about it lives in the numerator. These are the components MetricNet includes in operating expense, and the ones most internal calculations miss:
| Cost component | What belongs in it | Commonly forgotten |
|---|---|---|
| Direct agent cost | Salaries plus benefits, payroll taxes and bonuses for everyone who handles contacts | Benefits load, typically 20 to 30 percent on top of base salary |
| Indirect personnel | Team leads, supervisors, schedulers, dispatchers, QA reviewers, trainers, managers | The QA and workforce roles, which are usually charged to a different cost center |
| Technology and telecom | Ticketing licenses, telephony, ACD and IVR, knowledge base, chat, AI agents, laptops, headsets | Per seat license creep and the AI tooling added mid year |
| Facilities | Office space, utilities, insurance, allocated overhead | Nothing, if you are remote, but say so explicitly rather than silently dropping it |
| Travel, training and supplies | Onboarding and certification, offsites, materials | New hire ramp time, which is a cost even when it is not an invoice |
| Contractors and outsourced overflow | Any BPO, seasonal or agency spend for the same period | The overflow vendor used only in peak weeks |
Pull these from actual spend, not from budget. Budget tells you what somebody planned in October; your expense management system and general ledger tell you what really left the building last month, which is the only version finance will accept.
On the denominator, pick one definition of a ticket and hold it. Tickets resolved in the period is the standard, and it is better than tickets created because it matches cost to completed work. Whatever you choose, do not switch mid year, and exclude spam, duplicates and auto-closed noise. A ticketing system that auto-generates a case for every inbound newsletter reply will flatter your cost per ticket by 20 percent while nothing about the operation improved.
A worked example: what a 12 person support team costs per ticket
Numbers make this concrete. Take a US mid-market SaaS support team over one month:
| Line | Monthly cost | How it was derived |
|---|---|---|
| 10 agents, fully loaded | $68,000 | $65,000 average base, plus 25 percent benefits load, divided by 12 |
| 1 team lead and 0.5 QA | $13,000 | Same loaded basis, QA allocated at half time |
| 0.25 support manager | $3,300 | Manager splits time across support and success |
| Ticketing, chat and telephony licenses | $2,600 | 12 seats plus platform fee plus voice minutes |
| Knowledge base and AI assist tooling | $1,400 | Annual contracts divided by 12 |
| Facilities and overhead allocation | $4,200 | Finance allocation for 12 seats |
| Training, hiring and supplies | $1,500 | Trailing 12 month average |
| Total operating expense | $94,000 | |
| Tickets resolved | 5,200 | Excluding duplicates and auto-closed |
| Cost per ticket | $18.08 | $94,000 / 5,200 |
Notice what happens if you run the naive version. Agent salaries alone divided by tickets gives $13.08. That is a 28 percent understatement, and it is the number most teams quote when asked. It is also why in-house support so often looks cheaper than a vendor quote that includes management, QA and tooling by definition.
What is a good cost per ticket? 2026 benchmarks
There is no single industry average, and anyone quoting one figure without a source is guessing. The published ranges cluster like this:
| Source and scope | Reported cost per ticket | What it covers |
|---|---|---|
| MetricNet IT help desk benchmarking | $15.56 average, range $2.93 to $49.69 | Level 1 IT service desk, North America weighted |
| 2026 service desk benchmark reporting | About $22.50 per resolved incident | IT incidents including escalated resolution |
| The Office Gurus 2026 global blend | $8 to $12 blended, all channels and regions | Includes offshore delivery in the mix |
| Same report, North America only | $15.56 to $20 | Onshore delivery |
| Same report, Asia-Pacific delivery | $5 to $10 | Offshore delivery |
The spread is not measurement error. It reflects four things: where the agents sit, how complex the tickets are, how much of the volume the customer resolved without an agent, and whether the calculation was honest about indirect cost. Before you conclude that your $18 is bad, check that the benchmark you are comparing against used the same numerator.
What is the industry average cost per help desk ticket?
For a level 1 IT help desk in North America, the most widely cited figure is MetricNet's $15.56 average, with organizations ranging from roughly $3 to $50. Treat the average as a sanity check, not a target. A desk handling password resets and a desk handling clinical system outages should not have the same cost per ticket, and forcing convergence usually means underinvesting in the harder queue.
Cost per contact by channel
Customer support teams usually track cost per contact rather than cost per ticket, because a single case can span three interactions. The math is identical, but the denominator is contacts handled. Channel mix is the biggest single driver of the blended number:
| Channel | Typical 2026 cost per contact | Why it lands there |
|---|---|---|
| Phone | $9 to $16, over $20 for technical or white glove queues | One agent, one customer, no concurrency, longest handle time |
| Email or async ticket | $6 to $11 per case | Batchable, but multi-touch threads add hidden handle time |
| Live chat | $5 to $9 | Concurrency of two to three conversations spreads agent cost |
| Self-service resolution | $0.10 to $0.60, medians reported near $1.84 | Fixed content and platform cost spread across unlimited volume |
The gap between an assisted contact at a reported median of about $13.50 and a self-service resolution near $1.84 is the entire economic case for a knowledge base. It is also why ticket deflection rate is worth measuring as a money metric rather than a vanity one: every deflected contact is the difference between those two numbers, dropped straight to the bottom line.
How do you calculate cost per contact in a call center?
Take total contact center operating expense for the month, including loaded agent pay, supervision, telephony, workforce management tooling and allocated overhead, then divide by total contacts handled across all channels. Calculate it per channel as well as blended. The blended figure tracks your channel mix as much as your efficiency, so a blended number that improves while nothing operationally changed usually means customers simply moved to chat.
Why escalation is the most expensive thing your service desk does
Cost per ticket is not flat across support tiers, and the difference is larger than most teams assume. MetricNet's benchmarking consistently shows resolution cost climbing sharply at every level above the front line, driven by longer handle times and higher salaries: desktop or level 2 support typically costs multiples of the level 1 figure, and field or vendor resolution costs more again.
The part that gets missed is that escalation costs are cumulative, not substitutive. When a ticket is logged at level 1 and then escalated, you pay the level 1 cost and the level 2 cost for one outcome. A ticket that could have been closed on first contact and instead bounced through two tiers can cost three to five times the baseline.
This is what makes first contact resolution the highest-leverage cost metric in support. A five point FCR improvement removes the second and third contact entirely, which is worth far more than shaving 20 seconds off handle time.
How to reduce cost per ticket without wrecking the experience
Cost per ticket falls in exactly two ways: the numerator shrinks or the denominator changes shape. Cutting the numerator by cutting heads raises handle time, tanks service level and pushes contacts into a second attempt, which puts the cost right back. The durable levers work on ticket shape instead.
| Lever | What it changes | Realistic effect |
|---|---|---|
| Deflect the top ticket drivers | Moves high-volume, low-complexity contacts to self-service | Reported reductions of 85 to 95 percent on the automated categories |
| Raise first contact resolution | Removes repeat contacts and escalations on the same issue | Compounding, because it removes the cumulative tier cost |
| Fix the upstream process | Prevents the ticket from existing at all | The only lever with no ceiling |
| Improve routing accuracy | Stops the misroute and the reassignment loop | Cuts handle time and escalation rate together |
| Rebalance channel mix | Moves suitable volume from phone to async and chat | Directly lowers the blended cost per contact |
| Reduce agent turnover | Cuts hiring, training and ramp cost, all of which sit in the numerator | Often the largest hidden cost in a fast-growing team |
The third row deserves emphasis, because it is where operations work pays off and where most support cost programs never look. A large share of support volume is generated by something that went wrong earlier: a confusing invoice, a stalled onboarding step, a document that had to be resubmitted. Those tickets are cheapest to eliminate at the source. That is the same argument behind treating customer experience operations as a cost center you can actively design rather than a queue you can only staff.
Be careful with the metric itself. Cost per ticket is easy to improve by doing worse work: close cases faster, push customers to reopen, and the number drops. Always read it next to FCR, resolution time and CSAT. On its own it will reward exactly the behavior you do not want.
How to compare your cost per ticket against an outsourcing quote
This is the decision cost per ticket exists for, and it is where most comparisons go wrong. A BPO quote is usually priced per hour, per FTE or per contact. Your internal number is per ticket. Those are not the same unit, and converting between them is where the savings claim either survives or collapses.
Three adjustments make it apples to apples. First, convert the vendor price into your unit: a per-hour rate becomes a per-contact cost only once you apply their expected handle time and occupancy, not yours. Second, add the costs that stay with you after transition, which typically means vendor management, quality oversight, your ticketing licenses, and knowledge base maintenance. Third, model the ramp: a new vendor team resolves less per hour for the first 60 to 90 days, and that gap is real money.
Run that comparison against published customer service outsourcing costs and pricing before you commit to anything, and you will usually find the honest delta is narrower than the sales deck, though often still real for high-volume, low-complexity queues.
Where cost per ticket sits in your metric set
Treat cost per ticket as the financial layer over your operational metrics rather than a standalone KPI. Average handle time tells you how long work takes; cost per ticket prices it. Your staffing model tells you how many people the service level requires; cost per ticket tells you what that answer costs and whether the service level target is worth its price. Reviewed monthly alongside the rest of your customer service metrics, it turns support from a cost nobody can explain into a unit economic you can defend line by line. It also unlocks the next conversation, because once you can price a ticket you can attribute that cost back to the product lines and business units generating the volume, which is exactly the showback and chargeback model internal service teams use.
Frequently asked questions about cost per ticket
What is a good cost per ticket for a service desk?
For an onshore North American level 1 service desk, anything from roughly $15 to $22 per ticket is defensible depending on complexity and channel mix. Below $10 onshore usually means indirect costs were left out of the calculation, and above $30 usually means either genuinely complex work or a low deflection rate.
Should cost per ticket include the cost of self-service resolutions?
Include the platform cost in the numerator, but decide deliberately whether deflected sessions count in the denominator. Most teams report two figures: cost per assisted ticket and cost per total resolution including self-service. Reporting only the second makes deflection look like efficiency gains in the assisted channel, which it is not.
How often should you recalculate cost per ticket?
Monthly, from actual spend, with a rolling three month average for planning. It moves with hiring, license renewals and seasonality, so a single month read in isolation will mislead. Recalculate immediately after any headcount change or platform migration.
Does AI support automation actually lower cost per ticket?
It lowers cost for the ticket categories it fully resolves, where reported per-resolution costs fall to well under $3, but it raises the technology line in your numerator for every ticket. The net effect depends entirely on containment rate. If automation deflects 8 percent of volume while adding a platform fee across 100 percent of it, the blended cost per ticket can rise. Measure containment by category before and after, not overall.
What is the difference between cost per ticket and cost to serve?
Cost per ticket covers the support organization only. Cost to serve is broader and covers everything it takes to keep a customer running, including onboarding, billing operations, account management and success. Cost per ticket is one input into cost to serve, and it is usually the easiest one to measure first.