The in-house versus outsourced question is almost always argued on cost per hour, and cost per hour is the dimension where the answer matters least. Both arrangements can be made to produce roughly the same number. What differs is coverage, what happens when the person leaves, what you can see into, and whether your volume makes any of the expertise usable in the first place.
Three bodies of published data settle more of this than the usual spreadsheet does: the industry’s own compensation research, the platforms’ published volume thresholds, and the programmatic supply chain studies. None of them is about salaries.
The fee question has already moved
The historic argument for going in-house was that agency compensation scaled with spend rather than with work. That is now the minority arrangement.
What the survey found. The ANA’s compensation survey, 18th edition, fielded in the second quarter of 2022 across 101 marketers and 336 agency relationships, reported 82 percent of relationships as fee-based, up from 68 percent in 2016.
And incentives have receded too. Performance incentives were present in 41 percent of relationships, down from 48 percent in the prior edition and from 61 percent earlier.
What that changes about the comparison. If the external cost is a stated fee, the honest comparison is fee against fully loaded internal cost, not fee against salary. The spend-scaling objection has largely been answered by the market.
What it does not change. A fee tells you what you pay the agency. It does not tell you what the agency receives from anyone else, which is a separate question and the reason the disclosure question stays on the list either way.
And one caution about reading the survey. It samples large US advertisers. The direction of travel is informative for a small B2B buyer; the percentages are not a benchmark for one.
Before comparing arrangements, check whether either arrangement can produce the thing you are buying.
The social platform’s threshold. Roughly 50 optimization events per ad set within seven days of the last significant edit. Below that, the ad set stays in the learning phase and is flagged as learning limited.
The search platform’s published eligibility figures. Target return on ad spend requires at least 15 conversions in 30 days on search and shopping. Target cost per acquisition can begin with no conversion history, but Google recommends evaluating over 30 days including at least 30 conversions.
Divide before you decide. Monthly conversions divided by the number of ad sets or campaigns, compared against those figures. If the result is far below them, algorithmic optimization is not available to you at any staffing cost.
Which reframes the whole comparison. For a low-volume B2B account the value is in offer, targeting and creative judgement, not in daily account management. That judgement is bought in days per month, not in a full-time hire.
And it points the other way at scale. Once volume comfortably clears the thresholds and the work is continuous rather than campaign-shaped, the case for an internal function strengthens quickly, because the knowledge compounds where the data lives.
The cost lines the comparison usually omits
A salary is not a media function. Four other lines exist in both models and only one of them shrinks when you hire.
Tooling. Reporting, creative production, competitive research, call tracking, testing. These are subscriptions and they are the same subscriptions whoever operates them.
Creative production. The single biggest driver of paid performance is what the ad says and shows. An internal buyer without production capacity buys media for creative that somebody else has to make.
Coverage. An external supplier covers the calendar as a team. One internal hire covers it alone, minus statutory leave, public holidays, sickness and training. That gap is real and it lands in weeks where campaigns are running.
Measurement. Someone has to maintain the tracking, reconcile platform numbers with the CRM, and keep the reporting definitions stable. It is a distinct skill from buying and it is frequently assumed into the same job description.
What survives the hire. Only one of these four lines is genuinely replaced by a salary. The comparison that ignores the other three concludes in favour of in-house by construction.
Only one of five lines is actually replaced by an internal salary. The others exist in both arrangements. Source : Method (2026)
Transparency does not come with the employment contract
The strongest argument for in-house is usually visibility. It is a real gain, and it is narrower than it sounds.
What the supply chain study found. The ISBA and PwC programmatic study published in 2020 matched 31 million of 267 million impressions, about 12 percent, across the participating advertisers and their supply chains, and found roughly 15 percent of spend that could not be attributed to any identified party.
What that means for the decision. The unattributable portion sits between the buying platform and the publisher. Employing the buyer internally does not open that box, because the box belongs to the supply chain rather than to the buyer.
Where in-house genuinely wins on visibility. Inside the account. Every setting, every change, every historical comparison is yours, permanently, without an access request. That is worth having and it is the honest version of the claim.
And what the audit literature adds. ISBA’s audit guidance recommends annual audits in large markets, biennial in smaller ones, plus an exit audit at the end of a relationship, while noting that “access to vendor deal terms will be severely restricted due to confidentiality”. Even a formal audit does not reach everything.
The practical conclusion. If visibility is the reason, buy the visibility directly: account ownership, a dated change log, reporting that carries its definitions, and an exit audit clause. Those can be contracted for without changing who employs the buyer.
Framing the question as all or nothing hides the arrangement that works for most accounts below enterprise scale.
Keep ownership and measurement inside. The accounts, the tags, the datasets, the CRM reconciliation and the definitions used in reporting. These are the assets, and the platform rules make their location consequential.
Buy operation and creative outside. Bidding, structure, testing cadence and production. These benefit from seeing many accounts, and a single internal buyer sees one. That is the side of the split we are hired for, and our page on B2B paid acquisition sets out how bidding, creative and testing are run while the accounts and the datasets stay in your own portfolio.
Name an internal owner, not an internal buyer. One person, part-time, who owns the numbers, holds the definitions, reads the change log and asks the questions. That role is the difference between an agency being managed and an agency being trusted.
Why this is not a compromise. It puts each capability where it compounds. Knowledge of your buyers and your sales cycle stays where it is used every day. Knowledge of platform mechanics stays where it is refreshed across accounts.
And what it costs to get wrong in the other direction. The common failure is the reverse split: operation brought inside for cost reasons while measurement, ownership and definitions stay with a supplier. That produces an internal buyer who cannot verify their own results.
One test that tells you which side you are on. Ask who would notice first if conversion tracking broke tomorrow. If the answer is the agency, measurement is not inside yet, whatever the org chart says.
Answer these before running any cost comparison, because they determine whether the comparison is meaningful at all.
Is the work continuous or campaign-shaped? Continuous work with daily decisions rewards proximity and favours internal. Work that arrives in bursts leaves an internal hire idle between them.
Does your volume clear the platform thresholds? If not, you are buying judgement, not operation, and judgement is bought in days per month.
Does the knowledge compound in your business or in your category? Knowledge of your buyers, your offer and your sales cycle compounds internally. Knowledge of platform mechanics compounds faster across many accounts than inside one.
What happens the week the person leaves? Ask it in both directions. An agency loses a practitioner and reassigns. A one-person internal function loses the function.
And a fifth, which is really a decision rule. If you cannot answer the first four with your own data, the answer is not in-house yet. Build the measurement first, because an internal hire inherits whatever measurement you already have and rarely gets time to fix it.
Neither model wins in general. Each question points somewhere, and the pattern of the answers decides. Source : Method (2026)
Price the transition, not just the steady state
Whichever direction you move in, the first quarter costs more than either model does in isolation, and the platforms document why.
Relearning is not free. Automated bidding needs one to two conversion cycles to relearn after a change of conversion goal or action, and Google states that calibrating to a new objective can take up to three weeks. On Meta, changes to targeting, creative, optimization event or the addition of an ad are always-significant edits that restart the learning period.
Statistics do not travel with structure. Google states that when keywords are pasted into a new ad group, “the keywords’ performance statistics, such as impressions and clicks, are not transferred to the new location”. A rebuild loses the history against which you would have judged it.
Audiences have clocks. Google remarketing segments carry a maximum membership of 540 days and auto-close after 540 days of disuse; Customer Match requires at least 100 members; Meta website and app audiences carry 180 days. A transition that leaves audiences unused long enough starts them from zero.
Automation behaves differently under access change. Automated rules continue to run after access loss, while scripts stop, with Google stating that “the script will stop running but remain available”. Promotional credits are not transferred.
Which gives the transition a shape. Freeze structural change, hand over ownership and tags, verify tracking end to end, then let the relearning period run before comparing anything to the previous arrangement.
And a budgeting note. If you switch models and compare quarter one to the prior quarter, you are comparing a steady state to a relearning period. That comparison will favour whichever model you left, regardless of which is better.
What to do with this
Run the volume check before the cost comparison. Monthly conversions divided by ad sets, against roughly 50 per week per ad set on Meta and the published 15 to 30 conversions per 30 days on Google. That number decides whether you are buying operation or judgement, and the two are bought differently.
Then price the full external fee against the fully loaded internal cost including tooling, creative production, coverage and measurement upkeep. If the comparison still comes out close, the deciding factor is not cost. It is where the knowledge should live and what happens the week somebody leaves.
If visibility is the real motivation, contract for it instead. Own the accounts, require a dated change log, insist that reporting carries its definitions, and include an exit audit clause. The next steps are choosing a B2B acquisition agency and what agency reporting must contain.
Frequently asked questions
Is in-house media buying cheaper?
Not reliably, because the comparison usually leaves out tooling, creative production and coverage. And the fee question has moved: the ANA's 2022 compensation survey found 82 percent of agency relationships fee-based, up from 68 percent in 2016, so the external cost is increasingly a stated fee rather than a percentage of spend.
Does going in-house solve transparency?
Only partly. The ISBA and PwC 2020 programmatic study matched 31 million of 267 million impressions, about 12 percent, and found roughly 15 percent of spend unattributable. That opacity sits in the supply chain, not in the employment contract.
When does in-house clearly win?
When volume is high enough that the platform optimization thresholds are comfortably cleared, the work is continuous rather than campaign-shaped, and the knowledge being built compounds inside the business.