The comparison is nearly always framed as a salary against a retainer, which is the least useful way to look at it. One of those numbers is incomplete and the other buys a different shape of thing.
We are not going to quote a salary figure here. The surveys that circulate are inconsistent, rarely verifiable at source, and vary so much by market and seniority that a single number would mislead more than it informs. What follows is the structure of the decision instead, which transfers to whatever the salary actually is where you are.
The in-house cost nobody puts in the comparison
A salary is the start of the number, not the number. On top of it sit employer taxes and pension contributions, recruitment cost, software the person needs on day one, training to keep them current, the management time of whoever they report to, and the cost of the desk they occupy in whatever form your company pays for that.
Then there is ramp. A strong SEO hire is not productive in week one. They are learning your product, your buyers, your stack and your politics, and the technical ones are usually learning your codebase as well. Three to six months before meaningful output is normal and should be budgeted as part of the first-year cost.
And there is the capacity ceiling, which is the part that surprises people. One person covers one person’s week. SEO needs technical diagnosis, content strategy, editorial production and authority building, and the number of individuals genuinely strong at all four is very small. Most in-house hires are excellent at one or two and need help with the rest, which means the agency question returns anyway, in narrower form.
The agency costs nobody puts in either
The retainer is not the whole figure on this side either. Managing an agency takes real time from someone senior: briefing, reviewing, chasing, and translating between the agency and your engineering team. Budget a half-day a week and you will be close.
There is also the implementation gap. Some agencies ship changes directly; many deliver recommendations that then sit in your engineering backlog. A cheaper agency that only recommends is not cheaper if the work never lands — it is just a slower way of producing a document. Ask directly which one you are buying.
And context resets. An agency learns your situation slower than an employee does, and if the account team changes, some of that learning leaves. The mitigation is documentation, which is worth asking about explicitly rather than assuming.
The question that actually decides it
Ask whether your SEO work right now is foundational or operational.
Foundational means the big decisions have not been made: the architecture, the facet policy, the category map, the measurement model, the technical remediation. These are made once, early, and getting them wrong is expensive to undo. They benefit from people who have made them many times before, which argues for specialist help regardless of whether you also hire.
Operational means the decisions exist and the work is now execution and maintenance: shipping content against a map, keeping technical health, refreshing what decays, reporting. This is where in-house ownership wins clearly. It is cheaper over time, the context compounds, and the work benefits from someone who is in your product every day.
The sequence that wastes least, for most companies, is specialist help through the foundational phase, with documentation as a deliverable, then an in-house owner who inherits decisions rather than archaeology.
The hybrid that usually works best
An in-house owner plus specialist support is the most common mature arrangement, and it works because it splits along the natural seam. The internal person owns strategy, priorities, stakeholder management and institutional memory. The external team supplies the depth that is uneconomic to hire — technical diagnosis, content production at volume, digital PR — and the perspective of having seen the same problem elsewhere.
It fails in one predictable way: when the internal owner is junior and the agency effectively sets strategy while nobody internally can evaluate it. If you hire one person, hire the one who can judge the work, not the one who can produce the most of it.
One argument that has got stronger recently
The skill set has moved faster in the last two years than in the decade before. Ahrefs measured a 34.5% lower position-one click-through rate where an AI Overview appears. Research presented at ACM SIGKDD found that keyword density — a tactic many practitioners still carry — performed about 10% worse than making no changes at all on a live engine.
That raises the cost of a single person keeping current. Staying genuinely up to date is a meaningful part of a working week, and an individual absorbs that cost alone while an agency amortises it across clients. It is not a decisive argument, but it is a real one, and it has strengthened rather than weakened.
The counter-argument is just as real: an agency that is not keeping current is worse than an employee who is, and you cannot tell which you have from a pitch. Ask both the same question about what changed in the last six months and compare the specificity of the answers.