Media buying is the purchase of advertising space and time on behalf of an advertiser, and the management of its delivery. On the major digital platforms it is no longer a negotiation with a seller: it is participation in an auction.
That shift changes what the job is. You no longer agree a price in advance. You set targeting, creative and constraints, then observe the price that comes out, campaign after campaign.
This page covers what the work is, where the money actually goes, how the chain of intermediaries is built, the vocabulary you need to read a proposal, and what the available survey evidence says about how agencies bill.
Media buying and media planning are not the same job
They are usually held by the same team, which is why the distinction gets lost, and losing it is expensive.
Planning answers where and when. Which channels, which audiences, over which period, with what share of budget on each. It is a set of decisions made before any money moves.
Buying answers how much and how well. Executing that plan in the auctions, setting bids and constraints, watching delivery, and reallocating inside the plan as evidence arrives.
Why the confusion costs money. When results disappoint, the conversation almost always goes to buying: bids, structure, creative rotation. Sometimes that is right. Often the plan was wrong, meaning the channel or the audience never made sense, and no amount of execution fixes a plan.
The question that separates them in a review. Are we failing to reach the right people, or reaching them badly? The first is a planning problem. The second is a buying problem. They have different fixes and different people accountable.
Where the advertising money actually goes
There is one large US dataset with a declared method. Reading it properly changes what you expect from the market you are buying in.
The headline. IAB and PwC report US internet advertising revenue of $294.6 billion in 2025, up 13.9% year over year, in the thirtieth edition of a report running since 1996.
By format. Search $114.2 billion, display $81.6 billion, digital video $78.0 billion, digital audio $8.4 billion. Two cross-cutting figures that overlap those rather than adding to them: social at $117.7 billion, and programmatic at $162.4 billion, up 20.5%.
What the report is, and is not. It measures revenue booked by companies selling advertising. It contains no unit price of any kind, no CPM and no CPC. Anyone quoting it for a price is quoting it for something it does not contain.
What its own methodology admits. Data is “reported directly to PwC from companies selling advertising on the internet as well as publicly available corporate data”, non-participating companies receive “a conservative revenue estimate based on available public sources”, and the report states that “PwC does not audit the information and provides no opinion or other form of assurance.” The number of responding companies is not disclosed.
The line that changes how you read every market average. The top ten companies hold 84.1% of that revenue, up 3.4 points in a single year. The market described by the aggregate is, to a first approximation, ten companies. It is not the market a small advertiser buys in.
What that means practically. Growth rates from this report tell you about platform revenue, which is a decent proxy for competitive pressure in the auctions you enter. They tell you nothing about what you should be paying, and they should never be used to justify a budget.
How media buying actually works now
The mechanics matter because they determine what you can and cannot control, and most proposals describe them badly.
The auction, in one sentence. When a page or feed loads, an impression becomes available, eligible advertisers are evaluated, and one is served. The whole thing resolves in the time it takes the page to render.
Why the highest bid does not simply win. The platforms weight the bid against predicted engagement and ad quality. A more relevant ad can beat a higher bid, which is the single most useful fact in the mechanism: creative quality is a price lever, not just a response lever.
The chain, for everything that is not bought directly from a platform. A publisher offers inventory through a supply-side platform. An advertiser bids through a demand-side platform. The two meet on an ad exchange. Each intermediary takes a cut, and the cuts are not always visible to the buyer.
Why that chain is worth understanding even if you never touch it. It explains why a dollar of programmatic spend does not become a dollar of publisher revenue, and it is the structural reason competition authorities have taken an interest in the sector. If you buy display, it is your money moving through those hands.
What you control, and what you observe. You control targeting width, creative, constraints, and what you count as a conversion. You observe price, delivery and competition. Proposals that promise to control the second category are describing something that does not exist.
Which channels are bought, and which matter in B2B
The set of buyable channels is large. The set that repays a B2B budget is not.
Search. Buying intent that already exists. In B2B this is usually the first channel to work, because it reaches people at the moment they are looking, and the last to scale, because that moment is rare on a narrow market.
Paid social. Buying attention that does not yet exist. It reaches people who are not searching, which is most of your market at any moment, and it is judged on the wrong timescale more often than any other channel.
Professional networks. Precise firmographic targeting at a high unit cost. Worth it when the target list is genuinely narrow and the deal value carries the price, which is a narrower set of cases than the sales pitch implies.
Display and programmatic. Cheap reach with a long chain of intermediaries between your budget and the publisher. In B2B its most defensible use is retargeting, and its least defensible is broad awareness on a market of a few hundred companies.
Retail and commerce media. Growing fast and largely irrelevant unless you sell through those marketplaces.
The B2B filter to apply to all of them. Not “does this channel work”, which is unanswerable, but “can I reach a meaningful share of a few hundred companies here, and can I afford the unit cost of doing so”. Most channel debates end quickly once that question is asked out loud.
The vocabulary of media buying in eleven words
The jargon is short. These are the terms that recur, defined plainly. Keep this to hand when reading an agency proposal.
- CPM: cost per thousand impressions. What you pay for a thousand displays of your ad. On what it measures and what it ignores, see CPM.
- CPC: cost per click. What you pay each time someone clicks. You do not set it: it is an auction output, usually below your maximum bid.
- CTR: click-through rate. Clicks divided by impressions. The ratio rises when you narrow targeting, with no performance having changed: see click-through rate.
- CPL: cost per lead. What a contact costs. Two different objects share the name: see cost per lead.
- CPA: cost per acquisition. What a conversion costs, as the platform defines and attributes it. Not your customer acquisition cost: see how to calculate CAC.
- ROAS: revenue generated per unit of ad spend. A ROAS of 4 means four dollars of revenue per dollar spent.
- Auction: the mechanism that allocates an impression in real time, weighting bid against predicted engagement and quality.
- Inventory: the stock of advertising space a publisher puts up for sale.
- DSP: demand-side platform. The buying tool an advertiser bids through.
- SSP: supply-side platform. The selling tool a publisher offers inventory through.
- Ad exchange: the marketplace where the two meet.
How they connect. The publisher offers inventory through an SSP, the advertiser bids through a DSP, the two meet on an ad exchange through an auction, and the price is then read as CPM, CPC or CPA depending on what you bought.
How media buying agencies actually bill
The percentage ranges quoted across agency websites have no published survey behind them. The one recurring survey of the buyer side does, and it says something more interesting.
The evidence base. The ANA has run its Trends in Agency Compensation study every three years for close to fifty years among client-side marketers. The eighteenth edition was fielded in the second quarter of 2022 among 101 client-side marketers representing 336 agency relationships, an average of 3.33 relationships each.
Fees dominate, and increasingly so. 82% of respondents use a fee-based compensation model, against 68% in 2016. Among marketers spending $500 million or more a year, 53% now use fixed or output-based fees, up from 5% in 2016. Smaller advertisers lean toward labour-based fees, at 76%.
Performance pay is retreating, not advancing. Use of performance incentives fell to 41%, down from 48% in 2016 and 61% in 2013. The reported reason is blunt: most marketers said they do not know whether performance compensation is improving their agency’s performance.
Media buying is the exception. In the ANA’s separate 2019 media agency study, commissions remain far more common for buying than for other agency services: 24% of respondents use commissions for digital media buying, and 40% for programmatic media services, against 12% who reported commissions for agency services in general in the 2017 triennial survey.
Why that last figure matters to you. Commission on spend is the model most exposed to a conflict of interest, because the agency’s revenue grows with your budget rather than with your result. It survives in media buying specifically, and disproportionately in programmatic, which is also the least transparent part of the chain.
What the surveys do not give you. A fee scale. No institutional body publishes one, and the tiered percentage grids that circulate on agency blogs, twenty percent on the first tranche then fifteen then ten, rest on no published sample. Treat degressivity as a plausible negotiating principle and the specific tiers as folklore.
Which billing model fits your situation
None of the four is wrong. Each places the risk and the incentive somewhere different, and the right one depends on what you can measure.
| Your situation | What tends to fit | Why |
|---|---|---|
| Small spend, no reliable conversion tracking | Fixed retainer | Nothing else can be computed honestly, and percentage on a small budget prices the work below what it costs |
| Growing spend, clean tracking | Percentage, or hybrid | The work genuinely scales with the budget, and the tracking lets you check the result |
| Long sales cycle, hard-to-qualify leads | Avoid pure performance | Delayed conversions distort any cost-per-result figure, and the incentive drifts toward the easiest conversions |
| Large spend, mature team | Fixed or output-based fee | This is where the largest advertisers have moved, and it removes the spend-more incentive entirely |
Two cautions that apply to every model. A performance model is only as reliable as your conversion tracking; if conversions are mismeasured or delayed, the figure the fee rests on is wrong and the incentive points the wrong way. And no model substitutes for separating media spend from fees on the invoice. Ask to see both, separately, every month.
How do you know whether your media buying is profitable?
The question sounds simple and is answered wrongly almost everywhere.
What does not answer it. The conversions reported inside the platforms. Each attributes conversions to itself on its own window, which is why two platforms added together routinely claim more conversions than you actually received. We cover the mechanism in why GA4 and Meta conversions don’t match.
What does answer it. Total spend against deals closed, over a window that matches your sales cycle. If your deals close in six months, a campaign started in January shows up in June, and judging it in March measures a harvest during germination.
The metric hierarchy underneath. Delivery metrics like CPM and CTR describe the buying. Outcome metrics like CPA, ROAS and MER connect spend to revenue. Choosing which one you run on is a real decision, and we treat it in ROAS vs MER vs CAC vs LTV.
The honest limit. Attribution will not settle it either. Deciding whether a budget is genuinely producing incremental revenue requires an incrementality test, not a better attribution model, and most accounts are too small to run one cleanly.
What a monthly media buying report has to contain
Most reports are built to reassure. A useful one is built to let you disagree with it, and the difference is visible in about thirty seconds.
Media spend and fees, on separate lines. Not a total. If the two are blended you cannot compute a cost per result on the media, nor judge the fee against the work.
The delivery metrics, per campaign and per platform. Impressions, CPM, clicks, CTR. Aggregating platforms into one row produces figures that move for reasons nobody can name.
The outcome metrics, on the same rows. Conversions as defined in writing, cost per conversion, and where possible what happened to those conversions afterwards. A report that stops at the platform’s conversion count stops before the interesting part.
A dated change log. What was changed, when, and why. Without it, every explanation offered in the meeting is a reconstruction after the fact, and nobody can tell an intervention from a coincidence.
A comparison against the same month last year, not only against last month. Advertising inventory is seasonal, and month-over-month mostly measures the calendar.
What the report should not contain. A screenshot of the platform dashboard presented as analysis, a headline metric with no denominator next to it, and any benchmark quoted without its source and sample.
The tell that saves you an hour. Look at what sits on the first page. If it leads with impressions and CPM rather than with enquiries received and what they cost, the report is describing the buying rather than the result, and someone has decided which of those you should be looking at.
What loses money without anyone noticing
These are the leaks that do not show up as a bad number, which is exactly why they persist.
Budget spent on searches that were never yours. The default match settings serve queries related to your keywords, including ones that do not carry their direct meaning. On a narrow B2B market that is a steady outflow, and it is only visible in the search terms report. We cover the fix in Google Ads account structure for B2B.
Optimising toward the wrong event. If document downloads sit in your conversion set, the algorithm will conscientiously go and find more downloads. The account looks healthy and the pipeline stays empty.
Paying for the same person repeatedly. Overlapping audiences across ad sets means bidding against yourself, and the platform is under no obligation to tell you.
Creative wear that nobody dates. Performance decays gradually, and without a dated log of when each creative went live, the decline gets attributed to the audience, the season, or the algorithm. Creative fatigue shows up in CPM before it shows up in results.
Fees blended into media spend on the invoice. If you cannot see what was spent on media and what was paid in fees, separately, you cannot compute anything about either. This is a contract question, not an analytics question.
A reporting cadence faster than the sales cycle. Weekly readings on a small account measure noise, and reacting to noise restarts algorithmic learning each time, which costs real delivery.
Should you bring media buying in-house?
The answer is arithmetic before it is philosophical, and the arithmetic is usually skipped.
The case for outsourcing. Below a certain monthly spend, the fixed cost of a competent in-house buyer exceeds any fee you would pay an agency, and you get access to more accounts’ worth of pattern recognition than one person accumulates.
The case for in-housing. Above a certain spend the percentage becomes expensive relative to the hours, and proximity to the sales team is worth more than pattern recognition, particularly in B2B where qualification judgement is the scarce skill.
The question that decides it. How many hours a week does this actually take at your volume? Most B2B accounts at modest spend need a few focused hours weekly, which is neither a full-time job nor something to do in the gaps of another one.
The hybrid most people end up at. Strategy and qualification internal, execution external, with the conversion definition owned by whoever answers for revenue. It is unglamorous and it usually works.
One thing that does not change either way. Whoever runs it, you keep ownership of the ad accounts, the tracking setup and the historical data. An agency that runs your buying from its own account is holding your history hostage, and that becomes visible only on the day you leave.
Where to go next
You are structuring a search account. Google Ads account structure for B2B.
You run Meta. Meta Ads account structure and Meta Advantage+.
You are splitting a budget. How to split budget between Google and Meta in B2B, and small business advertising budget.
You want the unit costs explained. CPM, click-through rate and cost per lead.
You are weighing an agency against a hire. In-house vs agency B2B marketing, and the hidden cost of multiple agencies.
Your counters disagree. Why GA4 and Meta conversions don’t match and server-side tracking.
In short
- Media buying is now an auction, not a negotiation. You set targeting, creative and constraints, and observe the price.
- Planning and buying are different jobs. Ask whether you are failing to reach the right people, or reaching them badly.
- US internet advertising revenue reached $294.6 billion in 2025, up 13.9%, on a basis the report itself says is not audited and estimates for non-respondents.
- The top ten companies hold 84.1% of that revenue. A market average describes them, not your account.
- Fee-based compensation dominates at 82% of surveyed marketers, up from 68% in 2016, while performance incentives fell to 41% from 61% in 2013.
- Commission survives in media buying specifically: 24% for digital media buying and 40% for programmatic, against 12% for agency services in general.
- No institutional body publishes an agency fee scale. The tiered percentage grids on agency blogs rest on no published sample.
- Profitability is spend against closed deals, over a window matching your sales cycle, not platform-reported conversions.
Media buying is the operational core of paid acquisition, and it is judged on the wrong timescale more often than any other part of it. Book a diagnostic, or see how we approach B2B paid acquisition.