In-house, agency, freelancer and fractional are not better and worse. They fail in different directions: on resignation, on project close, when a calendar fills, and when you need parallel capacity. Choosing well means choosing the failure you can live with.
Most businesses do not choose between these four models. They fall into one, usually the one they used last time, and then spend two years discovering what it does not cover.
I have worked inside all four. I freelanced for years, I have been the engineer inside a business, I spent four months inside an agency, and I work fractionally now. That is not a boast, it is the reason I can describe how each one fails without guessing.
So here is the comparison written straight, including the parts that argue against hiring me.
Choosing wrong is not a disaster, it is a slow tax
Choosing wrong is rarely a disaster. It is a slow tax.
Hire in-house too early and you carry a salary through the quiet months and still have gaps, because one person cannot cover four disciplines. Hire an agency for something ongoing and you pay a premium for delivery capacity you stop needing the week after launch. Hire a freelancer for a system that has to keep working and you inherit it yourself by month nine. Take a retainer for a simple brochure site and you are buying attention nothing needs.
The models are not better and worse. They fail in different directions, and the trick is picking the failure you can live with.
The four models, side by side
| What you are comparing | In-house hire | Agency | Freelancer | Fractional engineer |
|---|---|---|---|---|
| Monthly cost | €9,200 to €9,600 loaded | €1,500 to €19,500 | Per project | €1,200 to €5,000 |
| Time to start | 5.4 months on average | Weeks | Days | Days |
| Disciplines covered | Usually one | Several, via specialists | Usually one | Four, one generalist |
| Capacity | One person, full time | High, parallel | Part of one person | Part of one person |
| Owns the outcome | Yes, while they stay | Until the project closes | Until handover | Continuously |
| Knowledge over time | Compounds, then leaves | Resets on rotation | Leaves with them | Compounds |
| Fails when | They resign | The project closes | Their calendar fills | You need parallel capacity |
| Best for | Technology is the product | Big, defined, finite work | Scoped jobs that end | Systems that must keep running |
What in-house actually costs
This is the number most comparisons quietly avoid, because the salary is not the cost.
A senior developer on a €90,000 salary in Germany costs €110,000 to €115,000 fully loaded once employer taxes, tools and management are counted. That is roughly €9,200 to €9,600 a month. Before they write a line, senior technical recruitment takes an average of 5.4 months.
And what you have bought is one person, in one discipline. Excellent at their thing. Not a server person and a data person and an automation person, because that is three or four salaries, which is why almost nobody under fifty staff has all of them.
None of that is an argument against hiring. It is an argument for hiring when technology is your product, because then you need that person there every day and the cost is obviously worth it. I have gone deeper into that arithmetic in fractional engineering vs hiring in-house.
The four, briefly
In-house wins when the technology is the business. Daily attention, full context, and someone who cares because it is their job and their equity. The failure mode is concentration: it all lives in one head, and that head can resign.
An agency wins when the work is big, defined and finite, and capacity is the constraint. Nobody else can put six people on your project for eight weeks. The failure mode is what happens after: ownership diffuses across a team that has moved on, which I have written about in fractional engineer vs agency.
A freelancer wins when the job has a real edge and finishes. Cheapest way to get a defined thing built well. The failure mode arrives about a week after launch, when the arrangement becomes “if they are free”, covered in fractional engineer vs freelancer.
A fractional engineer wins when the system has to keep working and nobody currently owns it. The failure mode is capacity: I cannot parallelise, and if you need four things built simultaneously I am the wrong answer.
Every one of these models works. They just fail in different directions, and most businesses pick by habit rather than by which failure they can afford.
The sequence that usually fits
For established businesses that are not technology companies, the pattern I see work is boring and rarely followed.
Use a freelancer or an agency to build the thing, sized to the job. Put someone on retaining it the day it launches, not the month it breaks. Hire in-house at the point where the work is genuinely full time and you can say what that person does every day for a year without inventing tasks.
The common mistake is skipping the middle step. The build gets a real budget and real attention, and then the system runs unattended until something goes wrong badly enough to notice, which is usually the most expensive possible moment to find out.
Before you weigh those four against each other, it helps to know what the thing you have now costs, which is rarely the number on the invoice. The cost calculator works it out.
Three questions, in this order
Three questions, in this order. The first one you can answer honestly decides it.
Is technology what you sell? If yes, hire in-house and stop reading. Everything else is a compromise you do not need to make.
Does this work end? If yes, and it is big, hire an agency. If yes, and it is contained, hire a freelancer.
If it silently stopped working, how long until someone noticed, and whose job would it be? If that question is uncomfortable, the problem is not who builds it. Nobody owns it, and a build does not fix that.
Everything else, including price, is downstream of those three.
Whichever model you choose, write down who is responsible for the system between requests. Not who built it, not who to email: who is looking when nobody has asked a question. If that line is blank, you have found the actual gap, and it is the same gap regardless of which of the four you hired.