In-house designer vs agency: the real math before you spend a dollar

A single in-house designer costs $60,000 to $100,000 a year in salary alone, before benefits, software, and the desk they sit at. A small in-house creative team runs north of $500,000 fully loaded, per Statista salary benchmarks for US creative roles. That is the leak nobody puts on the slide: you pay a fixed six-figure cost for an output that caps at roughly 4 finished assets a week from one person.

So the question is not in-house designer vs agency. It is which of four models matches your P&L and your output target. Most operators never run the comparison on a spreadsheet. They pattern-match to whatever their last company did, sign something, and discover the math eight months later when the renewal lands. Let me give you the math, then the honest fit, then who each one is wrong for.

The four options, not two

Most operators frame this as a binary. Hire someone, or rent an agency. There are four real options, and the fourth one barely existed two years ago. Each one wins on a different combination of volume, control, and cost structure. The mistake is not choosing wrong in the abstract. The mistake is choosing without knowing your own demand number first.

In-house designer

You get control, context, and someone in your standups. You also get a fixed cost that ignores demand. The hire takes 2 to 4 months. During that gap your pipeline still needs assets, so you bridge with freelancers anyway and pay twice. One creative director, working flat out, ships about 4 polished assets a week. When they take vacation, output goes to zero. There is no redundancy. A single designer also rarely owns motion, copy, web, and paid social at a high level, so you fill the gaps with contractors regardless. The fixed salary is the floor, not the ceiling, of your real spend.

Add the second-order costs that never make the offer letter. Benefits and payroll taxes add 25 to 40 percent on top of base, so an $80,000 designer is a $100,000-plus line item the day they start. The software stack - a Creative Cloud seat, a Figma seat, stock licensing, fonts - runs $2,000 to $4,000 a year per person. Recruiting burns weeks of a hiring manager's time, and a bad fit means a re-hire and another ramp. None of this argues against hiring. It argues for putting the full number on the slide first.

Freelancer

Cheapest per piece: $50 to $500 depending on scope. The catch is the hidden management tax. You write the brief, chase revisions, and re-explain your brand every single time. No redundancy either. When your best freelancer ghosts mid-launch, you start over from a cold brief. Good for spiky, occasional work. Bad for a steady content engine that has to feed paid channels every week without a gap.

The tax most operators underweight is their own time. If a marketing lead spends 4 hours a week briefing and reconciling freelance work, that is roughly 200 hours a year. Price that time at $75 an hour and the management overhead clears $15,000 - more than many people pay in freelance invoices. The freelancer line looks cheap because the expensive part is hidden inside a salary you already pay.

Agency

A campaign runs $2,000 to $10,000. A full brand package is $10,000 to $25,000. You get a team and a process. You also get the retainer, slow cycles, and the account manager who is a layer between you and the people doing the work. We wrote a whole teardown on this: the hidden cost of agency retainers is the slow bleed most mid-market operators never line-item.

The structural problem with the retainer is that it decouples your spend from your output. You pay the same in a month where the agency shipped three campaigns as in a month where they shipped one and spent the rest in alignment calls. The account manager exists to protect the agency's margin, not to make your assets faster. For a brand build with strategic weight, that process earns its price. For weekly paid-social volume, you are renting a Ferrari to drive to the mailbox.

The factory (the fourth option)

The factory is a codified system, not a person. You extract your Brand DNA once - 8 rules, 6 visual primitives, 1 voice doc - and the system ships 40+ on-brand assets a month. A lone creative director ships about 4 a week. That is a 10x throughput gap from the same brand spec. luup runs this as the Ad Factory in a 7-day sprint. The point is not cheaper labor. The point is that the brand rules live in the system, so output does not depend on one human's calendar or their notice period.

The thing that makes the factory work is the spec, not the speed. Once your visual primitives, your voice, and your hard rules are written down and machine-readable, the marginal cost of the next asset collapses toward zero. The first asset is expensive because you are defining the system. The fortieth is cheap because the system already knows what on-brand means. That is the opposite of the human model, where every asset costs about the same because each one starts in a person's head and ends in a review queue.

The 4-way comparison table

Here is the decision on one screen. Numbers are annualized or per-unit where noted.

Cost figures for salaries and team loads follow public US creative-role benchmarks via Statista. Agency and freelancer bands reflect standard market rates. The factory line is a structural claim, not a price quote, because the spend depends on volume.

A worked example: 30 assets a month

Pick a concrete number so the table stops being abstract. Say you need 30 on-brand assets a month to feed paid social, email, and landing pages. That is a realistic mid-market cadence once you are running more than one channel. Here is how each model performs against that target.

The in-house designer cannot reach 30. At 4 a week, the ceiling is about 16 to 17 in a good month with no vacation and no roadmap work pulling them off production. To hit 30 you hire a second designer, and your fixed cost doubles to roughly $200,000 loaded before shipping a single extra asset. Cost per asset lands around $550 at full output, and higher in any month where demand dips below capacity, because the salary bill does not flex.

The freelancer math looks tempting at first. Thirty pieces at an average of $200 is $6,000 a month in invoices. But you also need 30 briefs written, 30 review cycles run, and 30 brand reconciliations done by someone on your team. At 4 hours a week of management, you add roughly $1,200 a month in loaded internal time, and you carry execution risk every time a freelancer disappears. Call it $7,200 all-in, with no redundancy and a brand that drifts with every new contractor.

The agency quotes a retainer that, for 30 assets a month of mixed creative, typically sits in the $8,000 to $15,000 range, with volume above a soft cap triggering change orders. You get consistency and a team, but you wait on cycles and pay through the account manager layer whether the month was busy or slow.

The factory is the only model where 30 sits below the ceiling rather than above it. The system already ships 40-plus against a defined spec, so 30 is comfortable headroom. Your spend tracks the volume you use plus the one-time cost of the spec, and cost per asset keeps falling as volume rises, because the expensive work of defining the brand rules is already done. That is why the factory wins at high, steady volume and loses everywhere else.

How to actually decide

Run two numbers first. What is your real monthly asset demand, and what is the fully loaded cost per asset under each model? If you need 6 assets a month, a freelancer wins on math. If you need 40, the single in-house hire physically cannot produce them, and the agency will quote you into next quarter. The model that loses on a spreadsheet is the one you picked out of habit, not the one your numbers point to.

The fixed-cost trap is the one that kills mid-market budgets. A $500,000 team is a bet that demand stays high and steady. According to Harvard Business Review research on operating cost structure, fixed commitments amplify both your upside and your downside - they punish you hard when volume dips below plan, because the salary bill arrives whether or not the assets shipped. A codified system flexes with demand instead. You pay for the sprint and the volume you use, not for idle headcount in a slow month. That is the structural advantage, and it is the whole reason the factory model exists.

There is a second number worth running, and most teams skip it: speed to market. Teams that move faster on content compound their advantage, because more shots on goal means more learning per quarter. McKinsey research on marketing operating models points to the same pattern - the teams that win shorten the loop between idea and live asset. A model that takes two months to hire and another month to ramp spends its first quarter producing nothing. Factor that dead time in and the picture shifts again.

The build stack matters too. Agency website builds run 6 to 8 weeks. luup ships high-conversion sites in 7 days using Webflow, Framer, and Vercel, with design in Figma. Same logic: codify the patterns, ship faster, stop paying for discovery decks. Our 7-day launch breakdown shows the actual sprint, day by day.

Common mistakes operators make on this decision

The same five errors show up on almost every call. None of them are about the models. They are about how people reason under pressure when a budget question lands.

Treating headcount as the default

Hiring feels like progress. A new face in the standup signals to the rest of the team that marketing is serious. But a hire is the least reversible option on the list, and the slowest to deliver. If you do not know your monthly asset demand to within a rough band, hiring is a guess dressed up as a commitment.

Optimizing for the cheapest invoice instead of the lowest total cost

The freelancer invoice is the smallest number on the page, so it wins the room. It also hides the biggest variable cost on the list: your own team's time. Total cost per asset is the only honest comparison, and it has to include the hours your people spend briefing and reviewing.

Buying volume before defining the brand

You cannot scale output against rules that do not exist. Operators who jump straight to a high-volume model without a written spec get fast garbage, then blame the model. Define the spec first. The spec is what makes volume safe.

Paying agency rates for commodity volume

Brand strategy is worth agency money. Resizing the same campaign into nine ad formats is not. Mixing the two into one retainer means you pay strategic rates for production work that a codified system does at a fraction of the cost.

Ignoring redundancy until it bites

One designer or one freelancer is a single point of failure, and you only notice the day they are unavailable during a launch. Redundancy is not a luxury at scale. It is the difference between a system and a hostage situation.

What to ask before you buy anything

Before you sign with any of the four, get straight answers to five questions. The answers expose whether the model fits your reality or just sounds good in a pitch.

First, what is my real monthly asset demand across every channel that needs creative? Second, what is the fully loaded cost per asset under each model, including my own team's time? Third, what happens to my output the week the key person is unavailable? Fourth, how long until the first asset is live, and what am I paying during the ramp? Fifth, who owns the brand spec, and is it written down in a form that survives a personnel change? If a vendor cannot answer the fifth question, you are buying labor, not a system.

Who the factory is NOT for

The factory is wrong for you if your volume is low. If you ship 4 or 5 assets a month and they rarely change, a freelancer is cheaper and a system is overkill. Be honest about that number before you commit to anything.

It is also wrong if your brand has never been defined. You cannot codify rules that do not exist. Run the Brand DNA extractor first; if it comes back thin, you need a brand sprint before any factory makes sense.

And it is wrong if you genuinely need a person in your standups owning roadmap-level creative direction. That is an in-house hire, full stop. The factory ships volume against a defined spec. It does not replace a CMO's brain, and we will tell you so on the call rather than sell you the wrong thing.

It is also a poor fit if your work is one-off and bespoke by nature - a single hero film, a one-time rebrand, a piece of pitch theater that never repeats. A system pays off on repetition. When there is nothing to repeat, the setup cost has nowhere to amortize, and a specialist agency or senior freelancer serves that single high-stakes piece better.

See the case studies for the volume profiles where it does fit. If you are not sure which of the four you are, that is what the free Closed Loop Audit at /quiz is built to answer. When you want a human to walk the math with you line by line, contact us.