The three chairs I've sat in
Nearly every freelance developer vs agency comparison you can find was written by one of the sellers. Agency blogs conclude that you need an agency. Freelance marketplaces conclude that you need a freelancer. I'd like to be more useful than that, and I have one unusual qualification: I've been paid all three ways. Agency developer at FUSE and Envoc. Product and in-house developer at Trendsic and then Universe, where I'm now the web lead managing six developers and running the hiring pipeline. Freelancer under my own name since 1999. I've written the invoices, approved the invoices, and been the line item.
I list all that because it's the method. The answer to "should I hire a freelancer or an agency, or bring the work in-house?" changes with the job, and most people answering it professionally have only sat in one of the three chairs. The longer history is on my about page; what follows is each model described from the inside, including the one I currently sell.
Freelance developer vs agency vs in-house, at a glance
Here is the whole decision on one sheet. Every cell is my own read from twenty years of watching all three models succeed and fail from the inside, and the rest of this note adds back the texture the boxes flatten out.
| Factor | Freelancer | Agency | In-house |
|---|---|---|---|
| Cost | FreelancerOne senior rate, no markup | AgencyHighest hourly, blended across a team | In-houseSalary plus benefits, gear, and management time |
| Speed to start | FreelancerDays, if the calendar is open | AgencyWeeks of scoping and contracts | In-houseMonths of recruiting and onboarding |
| Breadth of skills | FreelancerOne strong generalist, deep in spots | AgencyWidest: strategy through QA under one roof | In-houseWhatever you hired for, narrowing over time |
| Continuity | FreelancerBus factor of one | AgencySurvives absences, but faces rotate | In-houseBest there is, until the resignation letter |
| Communication overhead | FreelancerNone, you talk to the person typing | AgencyHighest, a PM sits between you and the code | In-houseLow once the context soaks in |
| Best for | FreelancerScoped builds and specialist work | AgencyBig launches with many workstreams | In-houseProducts that ship every week for years |
Assessment: my own, from all three chairs · drafted July 2026, review before publish
What each one is actually good at
A freelancer is senior attention. Hire one directly and the person on the kickoff call, the person writing the code, and the person answering the Friday afternoon question are the same person. That is the product. Nothing gets lost in a handoff because there is no handoff. The model fits scoped builds (a marketing site, a landing page system, a component library) and specialist work (a performance rescue, an accessibility pass) where you need depth in one lane more than coverage of six.
An agency is coverage. Strategy, design, copy, development, QA, and project management under one contract is exactly what a big build with a hard date needs. When four workstreams have to land together in front of a stakeholder committee, the project management layer stops being overhead and becomes the thing you bought. My agency years left me with respect for that machine: nobody else ships a 200-page site, a rebrand, and a campaign in the same quarter.
An in-house developer is compounding context. Month one, they're learning the codebase. Year two, they know why the checkout is shaped that way, which promise created the strange banner logic, and what broke the last time anyone touched pricing. Context is the most expensive thing in software to rebuild, and employment is the only model that keeps it. When the work has no finish line (a product, a platform, a site that ships weekly), this is the chair to fill.
The honest failure modes
The freelancer's failure mode is the bus factor, and I say this as the bus. A solo developer has one calendar, one immune system, and one pair of hands; when I'm out, your project is out. One generalist is also the wrong shape for some jobs: a build that needs native mobile plus data engineering plus a full design team is not a search for a unicorn you should fund. The warning signs of an overbooked freelancer are slow replies, slipping dates, and discovery calls where you can tell you're the fourth priority of the day. Reference-check us the way you'd reference-check an employee; the hiring guide covers how.
The agency's failure mode is the switch after the pitch. The people who won your business are rarely the people in your repo, and your project's priority tracks your contract size; a small retainer at a big shop gets exactly the attention that math predicts. Questions travel through a project manager to reach the person who knows the answer, then travel back the same way, and the meter runs in both directions. None of that is malice. The economics of a bench work like that everywhere, and it's survivable if you insist on meeting the actual delivery team before you sign.
The in-house failure mode is slow and expensive. Recruiting takes months even when it goes well (I run a hiring pipeline, so this is not a guess), and a bad hire costs a year by the time you've noticed, coached, and replaced. Skills narrow, too. A developer four years into one codebase has been getting better at your product and quietly worse at everything else, which is fine right up until you need everything else. And the salary arrives in the slow months at the same rate as the busy ones.
Cost math that isn't apples to apples
Comparing the three on hourly rate is comparing apples to bowling balls, because the three numbers measure different things.
Start with the in-house developer vs agency cost question, because it's the one companies get wrong most often: they compare a salary against an invoice, and both numbers are lying a little. An employee's cost isn't the salary. The Bureau of Labor Statistics puts benefits at roughly 30 percent of total compensation (opens in a new tab) for private-industry workers, so a $120,000 developer actually runs closer to $170,000 a year before you've counted equipment, software seats, or the management time that keeps them pointed somewhere useful. In exchange, you keep every hour of context those dollars build.
- Salary
- $120,000
- Benefits load
- about $50,000
- What it really costs
- about $170,000
And the bar stops early. Equipment, software seats, and the management time that keeps the role pointed somewhere useful all sit outside it.
Benefits at roughly 30 percent of total compensation, per the BLS employer cost survey (opens in a new tab)
An agency invoice lies in the other direction. The hourly rate is the highest of the three because it's blended: inside it you're funding the project manager, the QA pass, the account layer, and the bench between projects. That money buys coordination a big build genuinely needs, and it's wasted on a build that doesn't.
A freelance rate reads high next to a salary until you notice what's missing from it: no benefits load, no bench, no minimum team size, no idle January. You buy the hours the work needs, then stop buying. The comparison that actually works is cost per outcome, not cost per hour: what does the launched thing cost each way, and who carries it afterward? I put real dollar ranges on common project types in the website cost note, and those totals move depending on which of these three models does the building.
How to decide in one paragraph
Here's the paragraph. Hire a freelancer when the work is scoped or specialized: you can describe the finish line, one senior generalist or one deep specialist covers the skills, and you want the person you talked to writing the code. Hire an agency when the work is wide and deadline-shaped: several disciplines, several workstreams, a date that can't move, and enough budget that the coordination layer earns its keep. Hire in-house when the work has no finish line and the product deserves compounding context. Mixing models is normal, not a failure: an agency for the launch, a freelancer for the specialist pass, an employee once the roadmap stretches past a year.
Most freelance developer vs agency advice ends with a pitch, so here's mine, priced honestly: the first chair is the one I sit in now. If your project is the scoped kind, this is the work I take on, and if it's one of the other two kinds, I'll tell you that on the first call. I've been all three. I know what each one is for.