Agency vs. in-house media buying: the real economics

Every company spending $30K+/mo on paid media eventually runs the same meeting: should we bring this in-house? The agency vs in-house media buying decision is usually argued with the wrong numbers on both sides. The in-house advocates count one salary against the agency retainer. The agency advocates count their awards. Both are selling. We run accounts for a living, so we have an obvious bias — and we also lose clients to well-executed in-house transitions, some of which we helped plan because it was the right call. This article is the math and the failure modes from both sides of that table, as honestly as we can write them.

The real cost of an in-house team

The number that anchors the meeting is a media buyer's salary: $85–130K for someone genuinely senior. The number that should anchor it is the cost of the function, because one buyer is not a media buying operation.

A minimum viable in-house team for $50–150K/mo of multi-channel spend:

• Senior media buyer — $90–130K base. Someone who has personally scaled accounts past your current spend level, not someone who "managed campaigns" inside a larger team where a director made the decisions. This distinction is the single most expensive thing to get wrong.
• Analyst / marketing engineer (often 0.5–1.0 FTE) — $70–110K. Tracking infrastructure, server-side events, attribution logic, reporting, incrementality tests. Without this role, the buyer optimizes against numbers nobody has verified.
• Creative production — $50–90K for an in-house editor/designer, or $3–8K/mo in contractor and UGC spend. Paid social lives or dies on 8–20 new creative units per month; a team without dedicated production starves within a quarter.

Loaded cost (benefits, payroll taxes, equipment, seat) runs 1.25–1.4x base. Add the tool stack the agency's fee used to absorb — analytics, creative tooling, spy tools, landing page software, typically $1,500–4,000/mo. Add recruiting: 20–25% of first-year salary per hire if you use a recruiter, and a realistic 3–5 months to fill the senior seat.

The honest range: $180–250K/yr loaded for the minimum team, before a single dollar of media. Against that, agency retainers for the same spend band run $36–120K/yr ($3–10K/mo, sometimes with a percentage-of-spend component). On raw line items, in-house costs 2–4x more than an agency at $50K/mo of spend. The multiple shrinks as spend grows: at $300K+/mo, the same team costs roughly the same while percentage-based agency fees keep climbing, and the math genuinely flips.

There is also a cost most spreadsheets omit on the agency side: management overhead. An agency is not zero internal effort. Someone senior on your side must own the relationship, supply context, review work, and make decisions — realistically 10–20% of a competent person's time. Count it.

What the spreadsheet cannot price

Two asymmetric risks sit outside the cost comparison, one against each model.

Against in-house: key-person risk. A one-buyer operation is a single point of failure holding undocumented knowledge of your account's history — what was tested, what died, why the structure looks the way it does. Senior buyer tenure in-house averages 18–30 months. When they leave, you do not lose a salary line; you lose the institution. We have taken over two accounts in the past three years specifically because a departed in-house buyer left behind an account nobody could explain, and both spent 2–3 months at degraded performance while the map was redrawn. Price this as the probability of departure times three months of performance drag — at $80K/mo of spend, that expected cost is not small.

Against agencies: attention dilution and misaligned defaults. Your account is one of 8–20 in a buyer's book at a typical agency. Percentage-of-spend fees quietly reward higher spend, not higher efficiency. And an agency's instinct under pressure is to protect the reported number, which is how accounts drift toward branded search and retargeting — activity that looks great and acquires nothing. These are structural tendencies, not universal behaviors, but you should select and contract against them explicitly: flat or performance-linked fees, named senior staffing in the contract, and full ownership of your ad accounts and data as a non-negotiable.

Where in-house wins

In-house is not the expensive option that vanity buys. There are configurations where it is simply correct:

• One massive channel. If 85% of your spend is Meta and it will stay that way, a dedicated senior Meta buyer living in your account 40 hours a week will usually out-execute an agency giving it eight. Depth beats breadth when breadth is not needed. This is the strongest in-house case that exists.
• Proprietary data as the edge. When your advantage is a feedback loop — CRM-qualified lead scoring flowing back into bidding, LTV models, margin-aware feeds — the work is continuous integration between internal systems and ad platforms. Internal teams sit next to the data and the engineers. Agencies bolt on from outside.
• Spend above ~$250–300K/mo. The fixed cost of a real team amortizes, percentage fees stop making sense, and you can afford redundancy (two buyers, not one), which retires most of the key-person risk.
• Creative as the moat. Brands whose organic content engine already produces winning material — some DTC and creator-led businesses — mainly need buying discipline attached to an existing creative machine, which is a narrower and more hireable job.

Where in-house fails

The failure patterns are as consistent as the success patterns:

1. The hiring problem is worse than expected. Genuinely senior buyers are scarce, expensive, and hard for a non-specialist to evaluate — the interview rewards confident vocabulary, and the market is full of people who ran ads inside a system someone else built. Companies routinely spend 4–6 months hiring, get a mid-level operator at a senior price, and discover it 2–3 quarters and $100K+ of mediocre spend later.
2. No testing infrastructure comes with the hire. A buyer without a testing system — hypothesis discipline, kill criteria, creative pipeline, incrementality checks — reverts to reacting to dashboards. Agencies amortize that infrastructure across a whole book of accounts; a solo hire has to build it from nothing while also running the account. Most never do. The account settles into maintenance: same structures, decaying creative, and a slow plateau that nobody flags because the weekly numbers move only 2% at a time. Diagnosing exactly that pattern is where disciplines like creative analytics earn their existence.
3. One brand, no cross-pollination. Platform behavior shifts constantly. An agency buyer sees a delivery change or a new format's real performance across twenty accounts within days; an in-house buyer sees one account and can't distinguish "the platform changed" from "something broke here." On fast-moving channels like Meta, this information asymmetry is worth real money in both directions — it is the honest core of the agency pitch, and it is true.
4. The function has no manager. If nobody senior internally can evaluate the buyer's work, in-house does not remove the oversight problem — it makes it worse, because there is no contract, no comparison set, and firing is slower than switching agencies.

Hybrid models, which is where this usually lands

Most companies we work with above $50K/mo end up in neither pure model. The stable hybrids:

• In-house owner + agency execution. A senior internal marketer owns strategy, budgets, and measurement; the agency runs the accounts. Fixes the oversight problem on the agency side and avoids the $200K team. The most common structure in the $50–200K/mo band, and the one we would recommend to a friend.
• In-house primary channel + agency secondary channels. The dedicated buyer runs the dominant channel; agencies run TikTok, Google, or experiments where 40 hours a week is not warranted. Rational, though it demands one internal owner of cross-channel measurement, or each party grades their own homework.
• In-house team + agency as auditor/sparring partner. A quarterly external review of a fully internal operation — structure, testing velocity, measurement honesty. Cheap ($2–5K per engagement) insurance against the plateau pattern, and roughly the shape of our audit.
• Agency now, planned transition later. Use an agency to build the system, hire the internal team against a working reference implementation, transition over 1–2 quarters with documentation as a contracted deliverable. Companies that hire into a functioning system succeed at rates that companies hiring into a vacuum do not approach.

A decision framework

Strip the vendors out of the room and answer five questions:

1. Spend level. Under $30K/mo: the in-house math almost never closes; agency or a fractional senior operator. $30–250K/mo: the contested zone — decide on the next four questions. Above $250–300K/mo: in-house core, probably, with specialists at the edges.
2. Channel concentration. One dominant, stable channel favors in-house depth. Three-plus active channels favor agency breadth or a hybrid.
3. Can you evaluate the hire? If nobody internally can distinguish a senior buyer from a confident mid-level one, you are not ready to hire one — whatever the spreadsheet says. Fix this first or rent the evaluation.
4. Where is your edge? Proprietary data and internal feedback loops argue in-house. Creative volume and cross-account platform knowledge argue agency.
5. What does failure cost? Model the 18-month total cost of each path including one realistic failure: a mis-hire and re-hire on one side, an agency switch and transition quarter on the other. The model with the cheaper failure mode is usually the right one at your stage, because over a long enough horizon you will experience the failure mode.

Run the numbers with loaded costs, priced risk, and management overhead on both sides, and the answer usually stops being ideological. It becomes a staffing decision with a spreadsheet behind it — which is all it ever was.