Hire an agency to run a channel that already works. Hire a fractional CMO to install the system that picks the channel. Hire a full-time CMO once that system runs.
That is the decision in three sentences. The rest is the evidence.
I have sat in all three seats. For fifteen years I was a VP of Marketing in PE-backed B2B SaaS. I was the full-time executive who owned the number. From that seat I hired agencies. I fired a few.
Today I run an operator-plus-platform model instead. So I am not neutral. But I will not sell you the option you do not need. The wrong one costs two quarters you cannot get back.
The three options in one table
Only the second column matters. Everything else follows from who owns the number.
| Option | Who owns the number | Typical monthly cost | Ramp time | When it fits |
|---|---|---|---|---|
| Marketing agency | You do. They own channel metrics only. | $1,000 to $20,000 retainer, plus media (WebFX) | 4 to 8 weeks to first deliverable | A proven channel needs volume |
| Fractional CMO | Shared. They own a named leading indicator. | $5,000 to $20,000 retainer, or $200 to $500 per hour | 1 to 4 weeks if they install | No demand engine, or nobody owns pipeline |
| Full-time CMO | Fully theirs, permanently. | $16,000 to $47,000 amortized, all in | 6 to 9 months hire plus ramp | The engine works and needs an owner |
The full-time figure is not a guess. Salary.com puts the average CMO base at $374,184 as of September 2026. The 25th percentile is $335,641.
Payscale samples smaller companies. It reports an average base of $191,521, with total pay reaching $353,000.
Take that gap seriously. The title spans a $10M ARR company and a $2B one. Plan against the figure that matches your stage.
Then add the real costs. Benefits and payroll tax add roughly 25 to 30 percent. Equity is on top. A recruiter takes 25 percent of first-year base.
Run the math. A $250,000 base CMO costs about $390,000 in year one. A $374,000 base CMO costs well over $560,000. Divide by twelve for the table above.
Fractional CMO vs marketing agency
The honest comparison is not skill. Plenty of agency operators are excellent. The comparison is structural incentive. Incentives beat intentions every time.
An agency is paid to run a channel. Its retainer renews when that channel looks healthy. So it optimizes what it controls. Cost per click. Cost per lead. Impressions. Rankings.
Those are real numbers. They are also the numbers least connected to revenue.
Watch what happens when the channel itself is the problem. Say paid search generates leads that never close. The correct call is to spend less on paid search.
An agency that makes that call shrinks its own retainer. I am not accusing anyone of bad faith. But you have hired someone whose income depends on the channel staying funded. Then you asked them to tell you when to defund it.
An agency cannot recommend less of itself. That is not a character flaw. It is a contract.
The second structural gap is the seam. Agencies work inside their channel. Nobody watches the handoff from paid into the CRM.
Nobody asks why one ICP segment converts at four times another. Nobody checks whether sales trusts the leads at all. Those failures live in the seams. No channel vendor owns a seam.
The third gap is the expensive one. When the contract ends, what do you keep? Ad accounts and creative files, usually.
The knowledge of why the account is built that way leaves with the agency. You rented capability instead of building it.
That is a fine trade for a channel you intend to rent forever. It is a bad trade for how your company generates pipeline.
A fractional CMO in the operator model is a different contract. The deliverable is a system installed in your stack.
A pipeline model. An attribution setup. A named leading indicator with a target. A weekly review your team runs alone.
If nothing is installed and handed off, you did not buy a fractional CMO. You bought consulting with a better title. The four-question test is in what a fractional CMO actually is.
There is a cost nobody prices in either. Managing an agency is a job. Someone briefs them, reviews the work, and chases the reporting. That is a real slice of a senior person's week.
I have seen founders spend six hours a month on agency management. Nobody puts that on the invoice. Add it before you compare the two retainers.
One more thing about agencies. They are the right answer more often than operator-model people admit. I say exactly when below.
Fractional CMO vs full-time CMO
This one is not about capability either. It is about time, and about what the seat is for.
Start with the clock. An executive marketing search runs six months. That is a healthy search, not a stalled one.
Then add 60 to 90 days before the new CMO changes a number. You are nine months out before the first metric moves. Meanwhile the board asks about pipeline every quarter.
Now the money. Use the Salary.com average base of $374,184. Add benefits, equity, and the recruiter fee. Year one lands north of $560,000. That is roughly $47,000 a month.
Even the Payscale average of $191,521 fits a smaller company at about $300,000 all in. Call it $25,000 a month. Either figure is real money spent before the engine exists.
Here is the trap I watched from inside PE-backed companies. The board pushes for a CMO because pipeline is unpredictable.
The company hires an impressive brand-side CMO from a much larger business. That person's real skill is running a marketing org that already works.
They arrive to no demand engine, no attribution, and three marketers. They build the org chart they know. Eighteen months later they are gone. Pipeline is still unpredictable.
The seat was wrong, not the person. A full-time CMO operates a functioning system. If the system does not exist, you hired an operator to be a builder. Different jobs.
Hire a full-time CMO to run a machine. Hire an operator to build one.
The inverse is also true. Most people in my category will not say it.
If your demand engine already works, hire the full-time CMO. If attribution is clean and pipeline forecasts within ten percent, hire full time.
If your only real bottleneck is that nobody owns marketing, hire full time. A fractional engagement is the wrong tool for permanent ownership. I will say that on the first call.
Every model priced out, including what I charge, is in what a fractional CMO costs.
When an agency is the right answer
I want to be specific. The operator-versus-agency framing gets lazy fast.
Hire an agency when all four of these are true.
One. The channel is proven. You can name a channel that produced closed revenue last quarter. And you can show the record, not a feeling. Proven means measured.
Two. The constraint is volume, not judgment. You know what to say and to whom. You just cannot produce enough of it. That is capacity, and agencies sell capacity well.
Three. Someone internal owns the number. A person on your payroll reviews pipeline weekly. That person can fire the agency for missing it. Without that owner, an agency becomes an unsupervised cost center inside 90 days.
Four. The craft is genuinely specialist. Production video. Technical SEO at scale. One ad platform's bidding mechanics. Deep skills that do not justify a full-time hire. Rent them.
Three of four is not enough. The one that fails most often is the third. It fails silently.
When a full-time CMO is the right answer
Four conditions again. They are stricter than most boards think.
One. The engine already runs. Predictable pipeline. Working attribution. A metric someone reviews weekly. You are hiring a driver, not a mechanic.
Two. The scope is genuinely permanent. A real team to manage. A brand to steward. A product marketing function. If the honest job fits in two days a week, the seat is not full time yet.
Three. You can absorb the ramp. Nine months to impact must not break the plan. If the board wants a new pipeline number in two quarters, this hire will not deliver it.
Four. You can compete on comp. At the median base above, plus equity, plus search costs. Without that budget you will hire a level below the seat. Then you will blame the person.
If you are PE-backed with an exit clock, the ramp condition usually fails. A hold period rarely has nine months of slack.
There is a sequencing answer that works, and boards under-use it. Bring in an operator to install the engine now. Hire the full-time CMO to run it in twelve months.
That order does two things. It moves a number this year instead of next. And it lets you write a real job description. By then you know what the seat requires.
The reverse order is what fails. Hiring permanence before the system exists is how good executives get set up to lose.
The fourth option: one senior operator plus an AI platform
There is a fourth shape, and it is the one I run. One senior operator does the judgment work. What to build, in what order, which number to move. An AI platform does the volume work underneath. Enrichment, reporting, attribution wiring, research. That work used to need three junior marketers or an agency retainer. So one operator can now install a demand engine and the measurement that proves it. The cost is a fraction of a full-time CMO, with no seam problem. The shared leverage is on the platform page. How the engagement runs is on the marketing leader page. That is the whole pitch. Back to the decision.
How to run the decision in a week
Five days. No consultants, no RFP. Do this yourself. The answer will be obvious by Friday.
Monday: write the number. One sentence. What number must change in two quarters, and what is it today?
If you cannot write that sentence, none of the three options fits yet. You need a diagnostic first. Stop here and fix that.
Tuesday: find the owner. Name the person on payroll who reviews that number weekly today.
If the answer is the founder by default, or nobody, an agency will not work. You are missing the seat above the agency.
Wednesday: audit the seams. Open the CRM and answer three questions.
Which source generated your last ten closed-won deals? What is lead-to-opportunity conversion by segment? Does sales act on marketing leads?
If those take longer than twenty minutes, you have a systems problem. A channel vendor will not fix a systems problem.
Thursday: price all three honestly. Agency retainer, plus media, plus internal management time. Fractional retainer for a defined window.
Then full time. Base, plus 30 percent, plus equity, plus a 25 percent recruiter fee. Divide by twelve. Multiply the ramp by nine months of no impact.
Put the three totals side by side. The full-time number is usually double what people carry in their head.
Friday: ask the ownership question. Ask each option one thing. What do I own when this ends?
An agency will describe ongoing management. That is honest. A full-time CMO will say everything. That is true.
A fractional operator should name artifacts. A pipeline model. An attribution setup. A documented operating review.
If the answer is a strategy or a roadmap, that is an advisor. Price it as advice.
Run those five days and you will not need my opinion. The evidence picks.
The short version
Agencies are capacity. Full-time CMOs are permanence. Fractional operators are installation.
Buying capacity when you needed installation is the most common mistake I see. It is expensive because the failure takes two quarters of channel data to prove.
Want to run the Wednesday audit with someone who has done it often? The growth diagnostic is thirty minutes and free.
If the question is timing rather than category, read when to hire a fractional CMO. Confused between operator roles instead? The fractional CMO vs GTM engineer piece covers that seam.
Sources
- Salary.com, Chief Marketing Officer benchmark: average base $374,184 as of September 2026, plus the percentile spread used above.
- Payscale, Chief Marketing Officer salary: average base $191,521 across 1,072 profiles. The smaller-company view.
- WebFX, digital marketing pricing: agency retainers $1,000 to $20,000 per month, with per-service ranges.
- Chief Outsiders, fractional CMO: a large firm's own framing of the category, including its agency comparison.
- CMOx, fractional CMO: the other major firm's packaging of the same role.