Comparing a Salary to an Invoice Is Not a Comparison

The conversation usually starts in a budget meeting. Someone notices the managed services line, does quick arithmetic, and observes that for roughly the same money the company could hire a full-time IT person who would be in the building every day.

The arithmetic is correct. The comparison is not, and the reason is that one of those two numbers includes a security stack, a documentation platform, after-hours coverage, vacation cover, and access to specialists, while the other number is a salary.

This is not a defense of managed services. Sometimes hiring is the right answer, and we have told companies so. It is an argument for running the comparison properly, because the version most companies run produces a decision that gets reversed eighteen months later at considerable expense.

What the salary number leaves out

Start with the obvious additions. Benefits and payroll burden add roughly a quarter to a third on top of base, depending on what you offer. Training and certification require a real budget; underfunding them leaves internal teams behind the environment they are supposed to run. Recruiting also has a cost, and turnover means paying it more than once.

Then the less obvious ones.

The tooling stack. Remote monitoring, endpoint detection, backup, email security, documentation, ticketing, and increasingly logging and identity tooling. Somebody pays for all of it regardless of who holds the contract, and a small business buys it at small-business pricing while a provider buys it at scale. When a company compares a salary to an invoice and forgets this line, the error is usually the largest single distortion in the analysis.

Coverage. One person covers about forty hours. If your business operates sixty hours because you have a warehouse, a clinic, or field crews, the internal model has a gap that either gets filled by paying someone to be on call, or gets filled by hoping. Both have a cost and only one of them appears in a budget.

Absence. Vacation, illness, and the two weeks in December. A provider absorbs this into a bench. A single internal hire cannot.

Depth. Your generalist will be genuinely good at some things and out of their depth in others, which is not a criticism, it is what generalist means. Security architecture, cloud migration, and incident response are specialist work, and the internal model buys them separately as projects. That line belongs in the comparison.

Key-person risk. This is the one nobody quantifies and the one that hurts most. When the internal IT person resigns, most of your institutional knowledge leaves with them, because documentation is a thing that happens when someone requires it and nobody requires it of an employee who is already busy.

What the invoice number leaves out

Fair is fair. The managed model has costs people also ignore.

Somebody inside your company still owns the relationship. The work may be part-time, but it requires an accountable person. Without that owner, the provider becomes a ticket queue and the strategic part of the arrangement disappears. We have watched this happen to good relationships more than once.

Onboarding is real work. Documentation, remediation of whatever was inherited, tooling deployment. It is a one-time cost that should be amortized honestly rather than pretended away.

Project work is usually outside the agreement. Migrations, new systems, and anything that changes rather than maintains the environment. A model that assumes the monthly fee covers everything will be wrong the first time you open a new location.

The third option most companies never model

Co-managed sits between the two and is frequently the right answer for a company between forty and two hundred people, particularly one with a line-of-business application that is central to how it operates.

The structure keeps one internal person who knows the business, the applications, and the people, and pairs them with a provider carrying the tooling, the after-hours coverage, the specialists, and the bench. The internal person stops being the whole IT function and starts being the person who translates between the business and the technology, which is usually what you actually wanted from them and rarely what their week allowed.

It fails in one specific way, and it fails this way often enough to name: nobody defines the boundary. Both sides assume the other has patching, or offboarding, or the backup verification, and the gap is only discovered during an incident. Co-managed works when the split is written down at a task level and reviewed quarterly. It does not work on goodwill.

How to run the comparison honestly

Put all three structures in the same model with the same lines. Every line appears for every option, even when the value is zero, because the zeros are informative.

Charge the tooling stack to all three. Charge internal coordination time to the managed option. Charge turnover and ramp to the internal options. Charge project budget to all three at whatever level your roadmap actually implies.

Then divide by something that means something. Cost per user per month is the only figure that compares like with like across structures, and cost per hour of business coverage is the figure that exposes whether a cheaper option is actually cheaper or simply covers fewer hours.

Be suspicious of any model, including ours, that arrives at a tidy answer. If two options land within about ten percent of each other, choose based on coverage, depth, and key-person risk because cost cannot distinguish them reliably.

The question underneath the question

Most companies that ask whether to hire or outsource are actually asking a different question they have not articulated: who is accountable for technology decisions here.

That question has a different answer than the staffing question. You can hire two people and still have nobody accountable for whether the technology investment connects to business outcomes. You can outsource everything and have excellent accountability, if the arrangement includes advice rather than only administration. The structures are independent, and confusing them is why companies switch models, get a slightly different flavor of the same frustration, and conclude the whole category is broken.

Decide the accountability question first. Then the staffing model becomes an operational choice about coverage and cost rather than a search for a solution to a problem it cannot solve.

The model

The workbook holds all three structures on the same basis, with every assumption visible and editable. It is deliberately conservative toward the internal option, because the internal option is the one most often understated. If managed or co-managed still wins on your numbers with those assumptions, the result is worth trusting.

The downtime section is labeled as an assumption rather than research, because we are not going to invent a benchmark. Replace those cells with your own incident history.

Compare the work before comparing the price

I start by writing the coverage model both options must provide. Who answers employees, manages identity, owns security operations, handles after-hours incidents, maintains the roadmap, delivers projects, covers vacation, and carries specialist work?

Once the work matches, the cost comparison becomes honest. A salary and a provider invoice can differ because they buy different coverage, or because one option carries waste. The model should make that visible.