Cost per ticket is everything your support desk costs you in a month, divided by the number of tickets it handled in that same month. That is the entire formula. You can run it on the back of an envelope inside twenty minutes, and further down we will go and do exactly that with real-shaped numbers so you can follow the working. So far, so harmless.
The catch is not in the arithmetic at all. It is that both halves of that fraction are choices dressed up as facts, and that the finished number behaves in a way nearly nobody expects, on account of it quietly rewarding the very things a good desk tries to stop doing. There is a spring, somewhere ahead of most support teams, in which the product gets better, the tickets get fewer, and this number climbs, and somebody in a meeting reads the climb as the desk going soft. This page exists mostly for that meeting.
Before any of it, the disclosure. Software is one of the lines in the numerator you are about to add up, and we sell some, so read the whole of this knowing where our bread is buttered.
What goes in the top of the fraction
Start with the people, since the people are most of it by a distance. Salaries for whoever answers, plus the employer costs that ride along on a salary. Those land somewhere between a fifth and a third on top depending on your country, and nobody ever remembers to include them. Then the awkward bit: the people who half-do support. The developer who loses three hours of a Thursday to the technical questions, the office manager who covers the inbox at lunch, the founder who still takes the angry ones personally. None of that appears on any budget line called support, and all of it is support cost, plainly said.
Then the tooling. The help desk subscription, whatever you pay for the mail itself, the phone system in case you run one, the screen-recording tool the team swears by, and any metered AI you have switched on. Training goes in, and the recruiting cost of the seat you refilled in March if you want the honest version of the year. The rest of it, rent and desks and the kettle, is a judgement call. Fair is fair, either judgement is defensible.
What is not defensible is changing the recipe between one month and the next. A cost per ticket that includes the developer’s Thursdays in January and forgets them by April has not measured a thing, it has only moved the furniture about. Pick a recipe you can stand behind, write down what went in it, and then leave the recipe alone for a good long while. Consistency does more for this number than completeness ever will, funnily enough, since you will mostly be comparing it against its own self month to month rather than against anybody else.
The bottom of the fraction has a smaller trap in it
A ticket wants defining before you count any. A conversation is one ticket, however many mails it took, and a customer who wrote four times about the one delivery is one ticket and not four, and any ticketing system worth the name counts it that way for you. Count raw mails instead and your denominator swells by half, your cost per ticket drops by a third, and the improvement is entirely fictional, on account of the desk having done not one thing differently and the money spent being the money that was always going to be spent.
The quieter leak is the work that never got counted at all. Every desk has some of it, no two ways about it: the question answered in a corridor, the reply sent from somebody’s personal address because the mail got forwarded there, the fix agreed over the phone and written down nowhere. All of that is real support, done by real salaried people whose salaries you just put in the numerator, and none of it lands in the denominator, which means your cost per ticket reads higher than the truth by however much of your work leaks around the record. Worth knowing which direction that error pushes before you go comparing your figure against anything.
The envelope, filled in
Round numbers, invented for the working, and your own will be different.
Say two people answer support full time at four thousand a month each in salary, and the employer costs bring the pair of them to somewhere near ten thousand. The developer’s occasional Thursdays are worth maybe another five hundred, honestly counted. Tooling, all in, comes to three hundred, and you decided against counting rent. That is ten thousand eight hundred for the month, the whole of it, near enough.
The desk closed eight hundred conversations in the same month. Divide the one by the other and you land at thirteen dollars and fifty cents a ticket. There is your number, and the whole exercise took less time than the meeting where somebody will eventually go and ask for it.
What gets quoted about the trade, for whatever hearsay is worth to you, runs from a few dollars a ticket on high-volume desks answering the same simple question all day, up past twenty on B2B desks where a single conversation can wander through three departments before it is done with. We would not lean on those figures for anything that matters, mind you, on account of every company cooking the numerator by a different recipe, which is rather the point of this page. Your thirteen fifty is worth having beside your own January figure, and it is worth very little indeed beside a number some other company arrived at by a recipe you never saw.
The spring the number turned on you
Now the mechanism, and it deserves a slow paragraph, because once you see it you cannot unsee it and most people have simply never had it pointed out.
Nearly everything in your numerator is fixed. The salaries do not shrink in a quiet month. The subscriptions bill the very same whether eight hundred conversations came through or five hundred did. So the top of your fraction mostly stands still, and the bottom of it moves about with volume, and a fraction with a fixed top and a moving bottom does something perverse: it gets better when you are busy and worse when you are not.
Follow that through one honest spring. In February somebody finally fixes the checkout fault that generated a fifth of your tickets, and the confusing dispatch mail gets rewritten so people stop asking where their order is, and the knowledge base picks up the four questions that made up another slice of the pile. Tickets fall from eight hundred to six hundred. Total cost has not moved an inch, the very same two people at the very same desks. So the sum now reads eighteen dollars a ticket where it read thirteen fifty at Christmas. The desk got more expensive per unit in the exact quarter the company got better at its job. And every one of those two hundred vanished tickets was a customer who never needed to be annoyed in the first place.
Read as an efficiency score, the number just told you a lie, and none of the lying had a thing to do with the desk its self. What it honestly measures is how full your capacity is, the very same way a half-empty bus costs more per passenger than a full one without the bus or the driver changing at all. A rising cost per ticket can mean a wasteful desk, or it can mean a desk sized for a volume that prevention work has gone and shrunk, and the figure by its self will never once tell you which of those two it is describing. Total support cost sitting beside it will. Keep the pair together, always, and be suspicious of anybody who reports the one without the other.
One average, wearing several different tickets
The other thing the single figure hides is the mix, and the mix is usually the interesting part.
A password reset and a billing dispute are not the same product. One takes ninety seconds and could be self-served entirely, the other takes three people and two days, and averaging the pair of them into one number produces a figure that describes neither. A desk whose mix drifts toward the hard kind, because the easy kind got soaked up by decent customer self service, will watch its average climb for the most flattering reason available: the cheap tickets left and the dear ones stayed, and the dear ones are dear because they are the work that actually needs people.
You do not need a taxonomy project to see this. Two piles will do, quick and slow, sorted by feel or by whether more than one person touched the thing. Go and cost each pile separately once, your own self, in one afternoon, and you will know your real numbers: perhaps three dollars for the quick kind against forty for the slow. From then on a shift in your average stops being a mystery, since you can check which pile grew before anybody starts drawing conclusions about the team.
The cheap ways down that cost you money later
Every lever that lowers this number quickly has a bill attached that arrives later, and it is worth knowing the shape of each before somebody pulls one in your name.
Closing tickets faster is the classic. Push the team to wrap things up and the minutes per ticket fall and so does the cost, and a share of those hurried closes come back open inside the week, each reopen being a second ticket that a slower first answer would never have created. The arithmetic of that trade is written up on the SLA compliance page from the clock side. The short of it is that the time never got saved at all, it only got moved a few days along, and it comes back with a second ticket attached to it.
Deflection theatre is the second lever. Put enough friction in front of the inbox and tickets genuinely fall, and the questions do not, and the difference between those two sentences is customers giving up on you quietly. Your cost per ticket improves while something you are not measuring gets worse, and the something eventually shows up over in your customer satisfaction metrics wearing a different name, a good while later, where it costs more to fix than it ever cost to answer.
And there is the tooling cut, which is the one we have the obvious interest in, so weigh this sentence accordingly: the record is usually the cheapest line in the numerator and the dearest one to lose, plainly said, on account of the out-of-band answering it prevents being invisible right up until it is everywhere. How that line gets priced in the first place, and what each pricing model quietly makes a team ration, is a separate argument we put on its own page as help desk pricing models.
Now the part where we sell you something
Maxdesk moves this arithmetic in one particular way, and we would rather describe it plainly than dress it up. We bill the workspace, not the people in it, so the software line in your numerator is flat: the free plan is $0 with unlimited people on it, Pro is $20 a month for the whole desk, Elite is $99 with the AI layer on top. Growing the team from three to eight changes your salary line and changes our line not at all, which on a per-seat product it certainly would, five times over.
The AI side we will describe with the brakes on. The AI customer support agent on Elite drafts and, where confidence is high, sends answers to the repeatable kind of ticket, which is exactly the cheap pile from the mix section above, and helpdesk automation routes and files so the humans start each conversation further along. What we will not print is a per-resolution saving figure, since we do not know your mix, your salaries or your recipe, and any number we put here would be an advertisement doing sums. Run the envelope your own self with the AI’s allowance and price in it and see what comes out, is the honest instruction.
Where our own numbers need the caveats read out
Three of them, in the interest of the whole page staying honest.
Elite’s AI carries an allowance, five thousand AI responses a month with paid packs past that, so with the AI switched on one line of your numerator is variable after all. It rises in your busy months the way the rest of our bill does not. Put it in the recipe as its own line. A metered cost hiding inside a flat one is exactly the kind of thing this page told you to go looking for, so we are pointing at our own.
Mail is the only door we watch. Desks built around a phone line, or around a chat widget, carry cost structures we never see and that the arithmetic above underestimates, since a phone queue staffs for its worst hour and mail staffs for its average day. If that is your desk, the envelope method still works, and our product is not the tool for you, and both of those things are true at once.
And the free plan keeps three rolling months of history, which matters here because the whole method runs on comparing months. Three points is a short trend. Twelve months of history rides on the $20 plan, twenty four on Elite, and if you would sooner not pay us, keep the monthly figure in a spreadsheet of your own and the method survives on any plan of anybody’s, which is the honest workaround and we would rather hand it to you than not. A small operation weighing the whole of this against its size has a separate piece to start from, customer service software for small business, aimed squarely at that reader.
Twenty minutes, one envelope, two recipes
Here is the exercise, and to be honest with you it is worth doing twice over with two different recipes, on account of the gap between the two answers being the actual finding.
First pass, the full recipe: people including the employer costs and the half-time helpers, tooling, training, your judgement call on overheads, divided by last month’s conversations out of your email management software or whatever holds the record. That is your fully loaded figure, the one that belongs beside total cost in any conversation about budgets.
Second pass, marginal only: strip it back to what one extra ticket actually costs you, which is a slice of somebody’s hour and near enough nothing else, since the salaries and the subscriptions were all being paid anyway. That figure is usually startlingly small, truth be told, single digits, sometimes cents, and it is the one that belongs in any conversation about whether to make contacting you harder. The first number tells you what the desk costs. The second tells you what an answer costs. What sits between the pair of them is your fixed base, the part that only structural decisions, head count, tooling, prevention, will ever move.
Most teams that run both sums find the argument they were about to have was aimed at the wrong number, and that the twenty minutes settled it before it started, which is about the best return on an envelope going.