I've been paid to buy media in just about every way a contract can dream up. CPA. Cost per sale — commission with a nicer haircut. Percentage of spend, which the big shops bolt on like seat warmers. Retainers. Hourly. I've sat inside all of them long enough to watch what each one quietly does to the work once the honeymoon ends and month eighteen rolls around.
Here's the part nobody says out loud at the pitch meeting: a compensation model isn't a price. It's an incentive wearing a price tag. It is never neutral. Whatever the model rewards is what you'll get more of — the scope document is mostly decoration. So the useful question isn't which model is virtuous. None of them is. The useful question is which model keeps its own incentive out in the open, where you can actually see it.
Hold onto that one. It's the whole argument. Everything below is just me showing my work.
The models, and their tells
Percentage of spend — my model, so I'll throw the first punch — pays the agency more as you spend more, efficient or not. CPA and cost-per-sale lean the other way: they tempt an operator to hoover up the cheapest, junkiest volume that still technically trips the conversion pixel. Every model has a tell. I'm not going to pretend mine was raised by nuns.
But here's the distinction that matters. Under percentage of spend, the rot is legible. You can see, to the dollar, that the fee is some slice of spend — call it ten to fifteen percent. The fuel-to-fee ratio sits right there on the invoice, where you can argue with it, cap it, or fire me over it. It scales with the work because spend itself is bounded; no sane client grows their budget to the moon, so the fee finds a ceiling alongside it.
And notice how rarely the industry actually picks the model you can watch. In the SoDA and Productive agency survey, commission-based pricing — the structure that ties an agency's pay directly to the spend it manages — came in at roughly 1% of agency revenue. Project work sat near half, retainers at 44%. Separately, the 4A's 2024 compensation survey pegged fixed fee as the primary model for about 72% of agencies.
~1%. That's the slice of agency revenue from commission-based pricing — the only model that welds an agency's pay to the spend it manages. The arrangement you can actually watch is the one almost nobody runs. The opaque ones are the house special.
Why the retainer feels clean — and isn't
Then there's the retainer. The crowd favourite.
Retainers are beloved because they're simple. One flat number a month. Clear in, clear out. What could possibly go sideways?
And in fairness, they're great — in the right room. A lawyer on retainer sells exactly one thing: their time and judgment. There's no second cost hiding behind the fee. The retainer is the product, full stop.
Media buying is a different animal, and that difference is the entire ballgame. A media buyer runs on fuel — the ad budget itself. That's the price of entry, the stuff that actually buys the clicks and the leads. The management fee is supposed to be a separate creature: payment for running the account, wiring up the tracking, optimizing, reporting, delivering the goods.
The retainer's quiet little sin is that it doesn't keep those two animals in separate cages. Fee and fuel get poured into one line on your P&L — "marketing," one number, once a month. This isn't a rare horror story, either: in AgencyAnalytics' 2024 data, retainers were the most popular package going, and most agencies fold reporting and management costs straight into the headline fee rather than itemizing them. And once fee and fuel share a line, guess which one wins. Not because anyone's twirling a moustache — because the structure makes the trade-off invisible. You simply can't see how much of your money turned into reach and how much turned into someone's margin.
What this looks like in the wild
I've walked into accounts that look exactly like this. Take an illustrative one, of the kind I've audited more than once: a client paying $3,000 a month for "marketing." Of that, roughly $2,000 is agency margin. Roughly $1,000 actually reaches the platforms. The numbers move around. The shape stays depressingly familiar.
Run the figure nobody handed the client: that's a 200% effective media fee. Two dollars of fee for every dollar of fuel. And it routinely gets worse — I've seen splits closer to 75/25, where three of every four dollars never leave the building. Here's the same idea as a picture: a clean, visible media fee against a bundled retainer.
Illustrative figures, drawn from the shape of engagements I've reviewed — not from the cited surveys. The point isn't the exact percentage; it's that the bundled invoice hides which slice is which.
Because media is the entry price — the literal thing that buys the result — starving it is a near-guarantee of a faceplant. Thin spend buys thin data, thin data buys lousy optimization, lousy optimization buys the gloomy little report that lands at month's end. The client's frustrated, and fairly so — they are paying real money. They just never got to see that most of it never made it into the tank.
I won't take work on these economics. Not purely on principle — they're rigged to fail, and I'd rather not sign up to own a result I was set up to lose.
The objection I should answer
By now a sharp reader has loaded the obvious rebuttal: well, you would say all this. You sell percentage of spend. This is a sales pitch in a lab coat.
Fair. So let me be precise about the claim. I'm not saying my model is pure — I took the first swing at it for a reason. I'm saying it's legible and bounded, and the retainer, in a media context, is neither. Under percentage of spend you can watch the fuel-to-fee ratio every single month and hold me to it. Under a bundled retainer you get one number and a story.
There's a second piece, and it's where the actual work lives: a good percentage-of-spend deal shouldn't reward raw volume either — that's my model's own trap. The fix is to feed quality signals back into the bidding so the spend chases value, not just more spend. But that's a whole other post, and I can hear my editor already.
The irony worth sitting with
A media buyer always works inside the gravity of your market and your platform. We can nudge CPCs, but we're playing between real floors and ceilings. We can't conjure demand that was never there.
What we can't do anything about is a structure that eats the budget before it ever reaches the auction.
So here's the irony I'll leave on the table. The same client who balks at a clean, visible ten-to-fifteen percent media fee will, under a tidy-looking retainer, happily pay an effective fee north of 200% — and never know it, because the number that would've told them was never printed.
The appetite for daylight is already there: the World Federation of Advertisers found that around 87% of marketers believe agencies resist more transparent fee models. The demand for the light switch exists. The wiring just isn't there.
That's the real danger of the retainer for anyone buying media. It isn't that the model lies to you. It's that it makes the single most important ratio in the relationship impossible to see. And you can't manage what you can't see.
Both hands on the wheel starts with knowing where the fuel is actually going.
Sources: revenue-share-by-model figures from the SoDA & Productive digital agency pricing survey; primary-compensation-model figure from the 4A's 2024 Compensation Methodologies Survey; package-type and fee-bundling data from AgencyAnalytics (2024); fee-transparency sentiment from the World Federation of Advertisers. Surveys measure different populations and define "retainer" differently, so prevalence figures vary across sources; the revenue-share framing is used here for consistency.
Want to See Where Your Fuel Is Actually Going?
Let's audit your campaigns and the economics behind them. No sales pitch — just honest analysis from someone who's managed this at scale.
Request a Campaign Audit