Agency versus in-house versus freelancer

The pitch decks all frame this as a cost comparison, and cost is the variable everyone gets wrong first, so it is worth clearing before anything else: cost-per-hour is close to the least important number in this decision. An agency's blended rate, a freelancer's day rate, a salary divided by working hours — line these up and you get a spreadsheet that feels rigorous and answers a question nobody should be asking. The real cost of each option shows up somewhere the spreadsheet doesn't look: in the hours someone senior spends managing the relationship, and in what gets lost every time the work changes hands.

That reframing isn't a dodge of the cost question. It's a better version of it. Once management time and context loss are priced the way payroll and invoices already are, the three options separate cleanly by the kind of work in front of you, not by which one is generically "best."

What a hire actually costs before anyone opens a laptop

Start with the number that's easiest to get wrong in the hire's favor: the fully loaded cost of an in-house role versus the sticker price of an agency retainer.

A salary is the smallest piece of an in-house hire's real cost. Add employer payroll taxes and benefits, typically 20 to 30 percent on top of base pay depending on jurisdiction and what the benefits package includes. Add recruiting: a role that takes six to twelve weeks to fill, during which the work either doesn't happen or gets absorbed by whoever's already stretched thin. Add ramp time once someone starts — a competent hire is rarely running at full output before the second or third month, because the first month is spent learning the product, the tools, the people and the parts of the job that were never written down. And add the downtime that recurs every time the role turns over: a hire who leaves after eighteen months takes their context with them, and the search-plus-ramp cost repeats.

An agency retainer, by contrast, is a number you can put in a spreadsheet cell and trust. It includes the agency's own overhead — account management, project management, the layer of people who exist specifically to make sure the specialist's time is used well — which is exactly what makes a blended rate look expensive next to a single freelancer's rate. That overhead is not waste. It's the thing that makes the retainer predictable in a way an in-house hire, with its variable ramp and turnover risk, structurally isn't.

Neither number is more honest than the other. They're honest about different things: the hire's number understates cost by omitting recruiting and ramp; the retainer's number overstates predictability by pricing an ongoing relationship as if it were a fixed monthly fee, when scope creep on either side is where retainers actually go wrong.

The hours nobody puts on a timesheet

Every one of these three options requires management time, and every proposal underprices it, because management time doesn't appear on the vendor's invoice or the recruiter's offer letter. It shows up as a hole in someone's calendar that used to be free.

An agency engagement that's actually moving needs a point of contact who can approve direction, answer questions the agency is blocked on, and sit through the weekly or biweekly status call. Five to ten hours a week is a realistic range for someone senior enough to make those calls, and if that person doesn't exist, the agency defaults to guessing, and guessing produces work that gets reworked, which is slower and more expensive than the retainer suggested.

A freelancer needs less structured management — there's no account layer translating your priorities into a team's task list — but more direct management, because you are the account layer now. Every question the freelancer would have asked an account manager, they ask you.

An in-house hire needs the least ongoing management of the three once ramped, which is the real argument for the option, not the salary math. A hire who has internalized the product, the codebase and the stakeholders can be handed ambiguous problems and trusted to resolve them without a weekly check-in, in a way that neither an agency nor a freelancer — both of whom are, by design, working from outside the organization — can match.

Option	Typical management load once running
Agency retainer	5-10 hours/week from a senior approver
Freelancer	Similar or higher — you are the account manager
In-house hire (ramped)	Lowest — delegation replaces oversight

What gets lost at every handoff

This is the cost the spreadsheet never has a line for, and it's the one that actually explains why some options are excellent for one shape of work and structurally bad for another.

Every time a project changes hands — from one agency to the next when a retainer ends, from a freelancer to whoever picks up after them, from one in-house hire to their replacement — the incoming party inherits the deliverables and none of the reasoning behind them. Why the homepage argument leads with a specific claim instead of the more obvious one, why a section was deliberately built and then deliberately left off the main navigation, which stakeholder vetoed which idea eighteen months ago and would veto it again — none of that lives in the file. It lived in the person, and the person is gone.

The cost of that loss shows up concretely on the second project, not the first. The first engagement with a new agency or freelancer usually goes fine, because everyone starts from scratch on a defined brief and a defined brief absorbs a lack of history reasonably well. The second engagement is where the tax gets paid: questions that were already answered get asked again, decisions that were already made get relitigated, and work that looks identical to what shipped eighteen months ago gets built as if it were new, because in every way that matters to the person building it, it is.

This is the actual argument for continuity, and it's a sharper one than "consistency matters." An in-house hire who's been there since the last project doesn't need the context rebuilt, because they never lost it. An agency handles this reasonably well within a single long-running retainer — the same team stays assigned, the same account manager carries the history forward — but poorly the moment the retainer ends and a different vendor picks up the account cold. A freelancer sits at the extreme on both ends: total continuity while they're engaged, since one person holds the entire project in their head, and total loss the moment they're not, since that context has nowhere else to live.

The freelancer's bus factor, and the contract clause that actually helps

The specific risk with a freelancer isn't quality — a good freelancer is frequently better, person for person, than the mid-level staff an agency assigns to a smaller retainer, because freelancing selects for people who can deliver without institutional scaffolding around them. The risk is concentration. One person holds all of the project's working context, and when that person is unavailable — sick, double-booked, unresponsive, or simply moved on to a different client — there is no backup, because a backup was never part of what you were paying for.

The mitigation isn't a personality trait to screen for in the hiring conversation. It's a contract term: documentation and handoff-ready files as a condition of each payment milestone, not as a courtesy the freelancer gets around to eventually. Source files, a short written rationale for the decisions that aren't self-evident from the file itself, and login credentials to anything set up on your behalf, delivered at defined checkpoints rather than requested after the relationship has already ended and the freelancer has less reason than ever to prioritize the request. This doesn't eliminate the bus-factor problem — one person can still become unreachable mid-milestone — but it bounds how much context is ever at risk of disappearing with them, and it's the difference between a freelancer relationship you can recover from and one you can't.

The verdict, by the kind of work in front of you

For a bounded project with a defined end — a site rebuild along the lines described in the redesign decision, a rebrand, a single campaign — an agency is close to the right default. The overhead that makes its rate look expensive is what buys predictability: a team that doesn't disappear if one person gets sick, a project manager absorbing some of the coordination load, and a natural endpoint where the retainer simply concludes rather than requiring anyone to make an awkward call about winding down a relationship. The brief matters more here than the vendor choice — see what a website brief should contain — because a well-scoped agency engagement is mostly downstream of a well-scoped brief, and a vague one produces the reworking cycle that erodes the retainer's apparent cost advantage.

For continuous work — a site, product or marketing function that needs steady attention for years rather than months — a single in-house hire is the reverse of the agency case in every respect that matters. The management overhead drops once the hire is ramped, in a way an agency retainer's overhead never does, because delegation is possible once someone has internalized the context and simply isn't possible with an outside vendor working from a brief. And the context-loss problem, which is the real cost center in continuous work, mostly disappears, because there's no handoff — the same person who built the thing last year is the one extending it this year.

For a specialist spike — an accessibility audit, a performance pass, a migration to a new platform, the kind of narrowly scoped expertise a generalist agency team or a full-time hire would either lack or be overqualified to maintain afterward — a freelancer specializing in exactly that problem is usually the sharpest tool available, provided the contract accounts for the bus-factor risk honestly. Nobody should be carrying a full-time accessibility specialist on payroll for a project that wraps in six weeks, and no agency's blended team matches a specialist who's solved this specific problem fifty times.

None of these is a permanent choice. The mistake worth naming, because it's the one that actually costs money — see what a redesign actually costs for what that number looks like once the hidden costs are included — is picking the option suited to a bounded project and then extending it, quarter after quarter, into the shape of continuous work it was never structured to hold, paying the context-loss tax every time the vendor or the contractor changes without ever admitting that the shape of the work changed first.