The question is usually framed as a monthly comparison: a retainer on one side, a salary on the other. That framing decides the answer before the analysis starts, and it decides it wrongly, because it ignores the two variables that actually move the number — fully loaded cost and utilisation.
Here is the comparison done properly: cost per delivered outcome, at three company sizes, with every assumption named.
First, fix the two numbers everyone gets wrong
Fully loaded cost, not salary
A USD 90,000 salary does not cost USD 90,000. Add employer taxes and statutory contributions, benefits, equipment, software seats, workspace, and the management time absorbed by having one more person. In most Western markets that lands at 1.25 to 1.4 times base. Use 1.3 as a working multiplier.
Then add the first-year cost of getting them: recruitment fees or internal hiring time, plus a ramp period where output is well below steady state. Two to four months of salary equivalent is realistic. It is a real cost even though it never appears on the org chart.
Utilisation, not headcount
This is the number that decides the whole argument. An agency specialist is billed across several clients, so you buy 0.2 of a paid media specialist, 0.15 of a developer and 0.3 of a designer. An employee is bought whole, whether you need them whole or not.
A performance marketer running USD 15,000 a month in ad spend is genuinely busy for perhaps one day a week. You still pay for five. The gap is not laziness; it is a mismatch between the shape of the work and the shape of employment.
Define the outcome before you compare
Cost per month is not comparable across models. Cost per outcome is. Pick a unit that both sides can be held to. In the tables below we use two:
- MQO — a marketing-qualified opportunity: a real conversation with a real buyer, accepted by sales.
- Release — one shipped, tested production change of meaningful size: a landing page system, an automation flow, an integration.
Your definitions may differ. What matters is that they are written down before the comparison, not after.
Company size 1: 20 employees
Assumptions: roughly USD 3M revenue, ad spend USD 8,000 per month, no marketing systems in place, founder still involved in every decision.
| In-house | Agency | Hybrid | |
|---|---|---|---|
| Structure | 1 marketing generalist | Full-service retainer | Founder owns strategy + agency executes |
| Annual cost | $117,000 $90k base × 1.3 | $96,000 $8k/mo | $78,000 $6.5k/mo + founder time |
| First-year hiring cost | $22,000 | — | — |
| Skill coverage | ~60% | ~90% | ~85% |
| MQOs per year (modelled) | 110 | 170 | 150 |
| Cost per MQO | $1,264 | $565 | $520 |
The verdict at 20 people: external, clearly. Not because agencies are better, but because you cannot hire one person who is credible at positioning, paid media, lifecycle email, analytics and web development at once. If you could, they would not take the role.
The failure mode at this size is hiring a junior generalist and expecting senior judgement. The second failure mode is buying a full-service retainer you cannot direct, which produces activity nobody asked for.
Company size 2: 200 employees
Assumptions: roughly USD 40M revenue, ad spend USD 45,000 per month, a CRM in place, a real sales team with targets, a product that ships regularly.
| In-house | Agency | Hybrid | |
|---|---|---|---|
| Structure | Head + 3 specialists + contractor budget | Retainer + media management fee | Head + ops manager, agency for paid, content and build |
| Annual internal cost | $611,000 | $52,000 0.3 FTE vendor management | $299,000 |
| Annual external cost | $60,000 | $396,000 | $216,000 |
| Total | $671,000 | $448,000 | $515,000 |
| MQOs per year (modelled) | 640 | 590 | 780 |
| Releases per year | 34 | 41 | 52 |
| Cost per MQO | $1,048 | $759 | $660 |
The verdict at 200 people: hybrid, and it is not close.
The reason is not cost, it is context. At this size the expensive scarce resource is judgement about your business: which segments matter, which claims are defensible, what the sales team will actually use. That must be internal. Execution capacity — media buying, content production, front-end build, data engineering — is a commodity you can rent at higher quality than you can hire at this scale.
Pure in-house at 200 people usually produces a team that is busy and under-specialised. Pure agency usually produces a vendor with better knowledge of your funnel than you have, which is a strategic risk regardless of how well they perform.
Company size 3: 1,000 employees
Assumptions: roughly USD 250M revenue, ad spend USD 220,000 per month, multiple product lines, a data team that already exists.
| In-house | Agency | Hybrid | |
|---|---|---|---|
| Structure | Team of 14 across strategy, performance, content, ops, creative | Two agencies plus a media buying partner | Team of 9 + specialist partners for creative, technical SEO, data engineering |
| Annual internal cost | $1,846,000 | $182,000 | $1,235,000 |
| Annual external cost | $180,000 | $1,584,000 | $540,000 |
| Total | $2,026,000 | $1,766,000 | $1,775,000 |
| MQOs per year (modelled) | 2,900 | 2,350 | 2,880 |
| Cost per MQO | $699 | $751 | $616 |
The verdict at 1,000 people: in-house for the recurring core, external for the spiky edges. Volume finally justifies full-time specialists, and utilisation stops being the problem it was at 20 employees.
What stays external even at this size: brand-level creative that benefits from outside perspective, technical specialisms you need four weeks a year, and anything where a single hire would create a single point of failure. The mistake at this scale is insourcing everything on principle and rebuilding a small, expensive agency inside the company — with worse tooling and no competitive pressure.
The pattern across all three sizes
| Headcount | Lowest cost per outcome | Why |
|---|---|---|
| Under 50 | Agency | Cannot achieve useful utilisation on any specialist role |
| 50–400 | Hybrid | Judgement must be internal; execution capacity is cheaper to rent |
| 400–800 | Hybrid, tilting in-house | Recurring work reaches full-time volume one function at a time |
| 800+ | In-house core, external edges | Volume justifies specialists; intermittent work still does not |
Notice that hybrid is never the worst option at any size. That is the practical takeaway. If you are unsure, the hybrid structure has the lowest downside across the widest range of conditions.
Five signals you have outgrown your current model
- Your agency knows your funnel better than you do. Time to hire an internal owner, not to change agency.
- Your internal specialist is a bottleneck on work they did not choose. Utilisation is above capacity in one skill and below 30 percent in three others.
- You are paying a retainer for work with no defined outcome. Either define the outcome or stop the line item.
- Briefs take longer to write than the work takes to do. That work belongs inside.
- Nobody can name last quarter's cost per outcome. Fix measurement before restructuring anything. See the data plumbing nobody budgets for.
What the models do not capture
Three honest limits on the maths above.
Quality variance inside each model dwarfs the difference between models. A strong agency beats a weak in-house team and vice versa, by more than any structural advantage. Structure sets your ceiling; the people you pick decide where inside it you land.
Institutional knowledge compounds internally and evaporates externally. Three years with the same in-house team produces context that no onboarding document transfers. Weight this if your product is complex or your sales cycle is long.
Switching costs are real. Moving models costs a quarter of reduced output regardless of direction. Do not restructure for a 10 percent modelled improvement.
If you want the same treatment applied to a specific programme rather than a whole function, the line-item approach in what a mid-market CRM implementation actually costs uses the same method. And if you are weighing whether we are the right external half of a hybrid, our services and recent work are the honest place to start.
Leave a Comment