Agency vs In-House Marketing: The Cheaper Option Flips Twice as You Grow

Agency vs In-House Marketing: The Cheaper Option Flips Twice as You Grow
The scale tips one way, then the other, then back. The chart is the part most cost comparisons leave out.

A median US marketing manager earns $166,790 a year, which works out to roughly $243,000 once benefits are added. That is about two years of a $10,000 monthly agency retainer. The cheaper option flips twice as a company grows: agencies win early, in-house wins once one channel runs every week, and hybrid wins after that.

Most agency-vs-in-house comparisons you'll find were written by agencies, and they reach the conclusion you'd expect. The in-house case usually comes from people selling in-house tooling. I went looking for numbers that neither side gets paid to publish (BLS compensation data, the SparkToro agency survey, the ANA and 4As tenure study) and built the comparison stage by stage. It's messier than either camp admits, which is probably why nobody sells it.

This is a zoom-in on one section of our guide to marketing team structure, where we argued that most companies end up hybrid. Here we put dollars on that.

The two numbers everyone gets wrong

The first mistake is comparing a salary to a retainer. A salary is maybe two-thirds of what an employee costs. The BLS employer cost data for June 2026 shows that for full-time private-industry workers, wages were 68.5% of total compensation and benefits the other 31.5%. Divide any salary by 0.685 and you get the loaded number. That's my arithmetic, not a BLS figure, and it still leaves out laptops, software seats, and the desk.

Run it on the BLS median and a $166,790 marketing manager costs about $243,500 a year. A $90,000 channel specialist costs about $131,000. A $60,000 coordinator lands near $88,000.

The second mistake goes the other direction, and agencies don't love talking about it: a retainer buys a share of someone's attention, and the share is smaller than most clients assume. Fluency's 2025 benchmark of 75+ US agencies and in-house teams found that account managers handle 35 clients on average, and agencies want to push that to 64. Thirty-five clients across a 160-hour month is about 4.6 hours of account manager time per client. Specialists and analysts sit behind that person, so the real hours are higher, but I'd still ask any agency to put the monthly hours by role in writing before signing.

What does a retainer actually cost? SparkToro's State of Digital Agencies survey, written up in January 2026, found 49% of agencies average $1,000 to $5,000 a month per client and 31% average $5,000 to $10,000. Its sample leans toward smaller shops, and the bigger retainers mostly show up in North America and Europe, so a US mid-market buyer should probably read the $5K to $10K band as typical.

Put those side by side and one loaded senior marketer equals two years of a $10,000 retainer, or four years of a $5,000 one. On paper, the agency wins easily. And sometimes it does.

Stage by stage: where the cost actually flips

The comparison changes with how much work you have and how varied it is. Here's how I'd price it at four stages. The dollar ranges come from the sources above; the stage thresholds are my judgment, so argue with them.

Stage 1: before a channel works (founder-led, under ~$10K/month in ad spend)

Agency or freelancers win, and it isn't close. You don't know which channel matters yet, so hiring a $131,000 loaded specialist means betting the salary on a guess. A $3,000 to $6,000 monthly retainer, or two freelancers on projects, buys you tests across several channels for under $75,000 a year.

The catch is that someone in-house still has to own the strategy, and at this stage that's the founder. If nobody inside can tell good agency work from bad, the cheap option gets expensive fast. We covered who to hire first, and when, in the first marketing hire decision.

Stage 2: one marketer, one channel starting to work

This is the first flip. Hybrid wins: one in-house owner plus agencies or contractors for specialist execution. Your first marketer at a $110,000 to $140,000 salary runs about $160,000 to $205,000 loaded, and maybe $60,000 to $100,000 a year goes to outside help.

Total: something like $220,000 to $300,000. More than a pure agency setup, yes. What you're buying is someone whose only client is you.

Recruiting time belongs in this math too. SHRM's latest recruiting benchmark puts the median time to fill a nonexecutive role at 39 days, and that's before ramp-up. If you start hiring in March, you probably don't get real output until June. An agency can usually start in two weeks. That gap is a cost, even if it never appears on an invoice.

Stage 3: one channel runs every single week

The second flip. Once a channel needs daily attention (paid social at $50,000+ a month, say, or a content program shipping weekly), bringing it in-house usually wins on cost and speed. A $131,000 loaded specialist who spends 40 hours a week on your account replaces a retainer that, at the Fluency ratio, may be buying 5 to 15 hours of combined attention.

The test I'd use: if a function would keep one person busy 30+ hours a week for the next 12 months, price the hire. If it wouldn't, keep paying the agency. Utilization is the whole game here. An in-house specialist at 50% utilization costs you double per useful hour, and that's the version of in-housing that ends in a layoff 18 months later.

Stage 4: a real team, many specialties

Hybrid wins again, for a different reason. A 10-person team can't economically employ every specialty it needs: video, CRO, programmatic, PR, data engineering. Each would sit idle half the time. So you keep the always-on channels in-house and rent the spiky ones.

This is roughly where large companies have landed. In The CMO Survey numbers we cited in the pillar guide, agencies and outside partners still do about a third of digital marketing work. Gartner's 2025 CMO Spend Survey found budgets split almost evenly between labor at 21.9% and agencies at 20.7%, with 39% of CMOs planning to cut agency allocations. Both lines are shrinking toward paid media, which took 30.6%.

The cheapest structure is almost never all-agency or all-in-house. It's whichever setup keeps every paid hour busy.

The costs that never show up in the comparison spreadsheet

Every cost model I've seen leaves out at least one of these. Anyway, here they are, roughly in order of how often they bite.

Your own management time. Somebody briefs the agency, reviews the work, sits in the weekly call. Call it 4 hours a week of a $243,500 loaded manager, which comes to about $117 an hour, or roughly $24,000 a year per agency relationship. Three agencies and you've quietly spent a third of a headcount on coordination. My guess is most teams underestimate this by half.

Switching costs. The ANA and 4As 2025 tenure study found client-agency relationships now average about seven years, up from 3.2 years in 2016, but media-only agencies average just 3.7 years. Big clients spend a reported $408,500 per pitch, per the same release. You won't spend that, but you will lose two or three months of momentum during any agency change, which in most cases I've seen is the bigger cost.

Markups and float. Media agencies earn on more than the retainer. Creator agencies keep a cut of influencer budgets that many brands can't see, and Meta's billing changes this year removed the float some agencies quietly lived on. Ask for the full fee structure, including anything paid by platforms or vendors.

Knowledge that walks out. This one cuts both ways. An agency leaves with your account history, and an employee leaves with your customer intuition. Small hedge here: I couldn't find good data on which loss costs more. My instinct is the employee, because agency work at least tends to be documented in reports.

How to run the comparison this afternoon

You can do this in a spreadsheet in about 45 minutes. One row per function (paid social, search, content, email, design, analytics):

  1. Hours needed per week, honestly estimated. If you don't know, look at the last three months of agency reports and count deliverables.
  2. Agency cost: monthly retainer × 12, plus any percentage-of-spend fees, plus $24,000 for your management time (adjust for your own loaded rate).
  3. In-house cost: salary ÷ 0.685, plus about $5,000 for tools and equipment, plus three months of salary as a one-time ramp cost in year one.
  4. Utilization: hours needed ÷ 40. Below 0.75, the in-house number is flattering itself. Multiply the in-house cost by (0.75 ÷ utilization) to see the real price of the idle time.

Anything where the adjusted in-house cost comes in under the agency cost, and utilization stays above 0.75 for a year, is a hiring candidate. Everything else stays outside. If you're doing this at budget season, it pairs well with the budget allocation framework we published last week, since agency spend is usually one of the hidden lines.

One caveat that I keep going back and forth on. AI tools are shifting the math in both directions at once. They let one in-house person cover more channels, which pushes utilization up. They also let agencies spread one account manager across more clients (the Fluency 35-to-64 target), which pushes your share of attention down. I don't think anyone knows yet which effect wins.

Common questions

Is a marketing agency cheaper than an in-house team?

At early stages, usually yes. A $5,000 to $10,000 monthly retainer costs $60,000 to $120,000 a year, against roughly $131,000 to $243,000 for one loaded US marketer. It stops being cheaper once a single channel needs 30+ hours of work a week, every week.

What does a hybrid marketing model cost?

For a company with one or two marketers, expect about $220,000 to $300,000 a year: the loaded salaries plus $60,000 to $100,000 in agency or freelance help. The point of hybrid is paying full-time rates only for full-time work.

How do I calculate the fully loaded cost of a marketing hire?

Divide the salary by 0.685, based on BLS data showing wages are 68.5% of full-time compensation costs. Add tools, recruiting, and about three months of ramp time in year one.

The spreadsheet is the easy part

My prediction: within 18 months, at least a third of the mid-market companies that in-housed a channel in 2025 and 2026 will move part of it back out, mostly because they hired for a peak workload that didn't last. In-housing itself was probably fine for most of them. They just never checked utilization after the first quarter.

So whatever you decide, put a date on the calendar six months out to rerun the four steps. The right answer at 8 people probably isn't the right answer at 15, and the companies that overpay are mostly the ones still running last year's structure.