A tech gives notice. The budget review that follows usually goes one of two ways: don't backfill the seat, or go down the vendor invoices and cancel a couple of tools. Both feel like IT cost reduction. Neither touches where most of a service desk's money goes.<\/p>\n
Most "IT cost reduction" advice is written for a CIO with a data center and a software sprawl problem: consolidate vendors, rightsize cloud, audit licenses, renegotiate contracts. That's fine advice for that reader. For an MSP owner or service manager, the biggest line is people, and the way to spend less on people without losing any is to stop paying them to do work that isn't fixing things.<\/p>\n
The short answer:<\/strong> for an MSP service desk, IT cost reduction means lowering your cost per ticket. You do that by removing the minutes around each fix (writing notes after the call, chasing missing details, touching duplicate tickets, escalating work Level 1 could have closed, waking the on-call tech for something that could wait) and by billing the time your techs already work. Cutting headcount moves the same tickets onto fewer people. It doesn't make them cheaper.<\/p>\n The cleanest way to see your cost is the metric benchmarking firm MetricNet has used for years: cost per ticket is your total monthly operating expense for support divided by your monthly ticket volume<\/strong>. Operating expense includes technician and supervisor salaries and benefits, technology and telecom, facilities, travel, training and supplies (HDI, Jeff Rumburg, "Desktop Support Cost per Ticket," 2017<\/a>).<\/p>\n When MetricNet broke that expense down for desktop support groups, technician salaries and benefits were more than 60 percent of the total, and about 74 percent was personnel once you add the non-technician staff (HDI, "Technician Utilization for Desktop Support"<\/a>). Those figures are about a decade old and come from internal IT groups, not MSPs, so don't quote them to your board as your number. The shape is what matters, and you can check it against your own P&L in five minutes: put a month of payroll for your service desk next to a month of software invoices. For most MSPs it isn't close.<\/p>\n That's why canceling tools rarely moves the needle. A license you cut is a real saving, but it's a small one next to the hours your team spends every day on work around the work.<\/p>\n If you don't backfill a seat, your ticket volume stays the same. Clients still call, alerts still fire, onboarding still happens. What changes is how many people absorb it.<\/p>\n MetricNet calls technician utilization "the biggest lever" for controlling support cost and puts the average for desktop support groups around 57 percent, with a range from 26 to 89 percent (HDI<\/a>). Running fewer techs harder does raise utilization. It also shows up as slower first response, more tickets bounced to senior techs because the junior ones are buried, and more notes written from memory at 5:30. You traded a salary line for a quality problem, and the quality problem has its own cost.<\/p>\n The better question is: of the hours you already pay for, how many go to fixing things?<\/p>\n These are the spots where a service desk spends time that isn't resolution. None of them needs a new PSA, and each can be measured in the PSA you already run, whether that's ConnectWise PSA, Autotask, HaloPSA or Syncro.<\/p>\n A tech spends twelve minutes on the phone fixing an Outlook profile, then spends a few more writing the ticket note and the time entry. Or they don't, and write it at the end of the day from memory, or never. Either you pay for the writing time or you lose the time entry and the history.<\/p>\n This is the easiest leak to close because the call itself already holds the note. Vight<\/a> pulls each recorded call from the phone system (3CX, Teams, RingCentral, Zoom Phone, Dialpad and others), transcribes it, matches it to the client and ticket, and posts a summary of the issue, the steps taken and the resolution as a ticket note with a time entry. The tech takes the call the way they do today. We wrote more about why this is where time tracking breaks in MSP time tracking<\/a>.<\/p>\n "Printer not working" from a user at a client with nine printers. The first touch on that ticket isn't a fix. It's an email asking which printer, what the error says and when it started, then waiting for the reply.<\/p>\n Write down, per type of request, what a tech needs before they can work it, and get that information at intake instead of at first touch. Vight's ticket triage<\/a> matches each new ticket to a scenario you configure, emails the requester for the details they left out, and posts an internal note saying what's answered and what still needs confirming. If you're building this by hand first, our ticket triage guide<\/a> covers the scenarios worth starting with.<\/p>\n When the internet drops at a 40-person client, you don't get one ticket. You get one from each person who emails the portal, plus a phone call or two. Each one gets opened, read and closed by someone.<\/p>\n Merging them is cheap time back. Vight's deduplication agent compares new tickets within the same client, by exact title and then by AI similarity, closes the duplicate with an internal note and leaves the parent open. It runs on ConnectWise, Autotask, HaloPSA and Syncro.<\/p>\n Every tier a ticket climbs adds cost. MetricNet's older North American figures put an average ticket at about $22 at the service desk, $62 at desktop support, $85 at Level 3 and $196 for field support, and the costs stack: a ticket that goes from the desk to desktop support costs both (HDI, "Percent Resolved Level 1 Capable," 2011<\/a>). The dollar amounts are dated. The ratio between tiers is the point.<\/p>\n The same research found that on average 18.6 percent of tickets resolved by desktop support could have been resolved at Level 1, and that organizations with a strict single point of contact averaged 15.3 percent against 22.8 percent for those without. At an MSP, the usual reason a Level 1 tech escalates a known issue is that the fix lives in a senior tech's head or in a ticket from eight months ago nobody can find.<\/p>\n Make past fixes findable. Ask Vight<\/a> answers questions over your calls, tickets and documentation (IT Glue, Hudu, SharePoint) and cites the ticket, call or document each answer came from, inside the ConnectWise, HaloPSA and Autotask ticket pods and in Teams and Slack. Track your own escalation rate before and after, by client and by request type.<\/p>\n Your on-call tech gets a 2 a.m. call about a password reset that could wait until 8. That's a cost even if you don't pay a call-out fee: it's the tech who's slower the next morning and the one who eventually quits over on-call.<\/p>\n Separate urgent from not urgent before a person is woken up. Vight's AI voice agent answers after-hours calls on a Vight phone number, captures the caller's details and opens a ticket with the context, and the after-hours escalation agent pages the on-call tech only for urgent ones. If you're deciding whether to bill for after-hours work at all, read should MSPs charge for after-hours support<\/a>.<\/p>\n This one isn't a cost cut. It's revenue you already earned. On block-hour and time-and-materials agreements, a fix with no time entry is free work. On flat-fee agreements, missing time hides which clients cost more to serve than they pay.<\/p>\n Vight's Time Review<\/a> puts a day of each technician's time entries on a timeline next to their calls and work notes and flags ticket work with no time logged and entries that overlap. Profitability puts invoices next to logged labor cost, client by client. More on this in MSP profit margins<\/a> and MSP billing increments<\/a>.<\/p>\n You don't need a consultant. Pull these from your PSA for last month, then again 30 days after you change something:<\/p>\n\nWhere service desk cost actually sits<\/h2>\n
Why cutting a tech usually doesn't cut cost per ticket<\/h2>\n
Six places the minutes leak<\/h2>\n
1. Documentation after the call<\/h3>\n
2. Tickets that arrive missing the facts<\/h3>\n
3. Duplicate tickets<\/h3>\n
4. Escalations Level 1 could have closed<\/h3>\n
5. After-hours calls that didn't need a human<\/h3>\n
6. Time worked and never billed<\/h3>\n
How to measure it in 30 days<\/h2>\n