I sat through two agency pitches for my own brand and ran both quotes against my account. A management fee should be measured against what advertising actually contributes — the gap between your break-even ACoS and your real one — not against revenue. On my numbers that gap is 9.6 cents per dollar of ad sales.
I have never fired an Amazon PPC agency. I’ve never hired one either. I sat through two pitches for Rowdy Rooster Woodworks, took both quotes home, and ran them against my own account. I didn’t sign with either one.
So take this for what it is — not a breakup story, but the arithmetic I did on the buying side of the table. If you already have an agency, it’s the same arithmetic. You just have invoices instead of a quote.
The bias is obvious: I sell software that does part of what an agency does. The numbers below are from my account, and the method is one you can run on yours.
The structure isn’t the point
Agencies price differently and there is no single norm worth quoting at you. Some charge a percentage of ad spend. Some charge a percentage of ad-attributed sales. Some charge a flat management fee, and plenty charge a base plus a percentage. A percentage of ad spend has started showing up in software pricing too, not just agency pricing.
I was quoted two of those shapes. One was $2,000 to set up, then $500 a month in software fees, plus 5% of ad-attributed sales. The other was a $2,000 monthly base that would move to 5% of ad-attributed sales once the percentage outgrew the base. I’ve written up both pitches in full elsewhere, including the one that presented fabricated Amazon listings as case studies.
But I don’t want this to rest on two pitches I happened to sit through. Whatever the structure, there is one test, and it’s the same test every time.
A percentage of sales is not a percentage of anything you keep
Last 30 days on my US account: $5,070 in ad spend against $14,164 in ad-attributed sales. That’s a 35.8% ACoS.
My break-even ACoS — the point where a sale stops making money — runs from 28.9% to 50.8% depending on the product. Weighted by what actually sells through ads, it comes to 45.4%.
So what advertising contributes is the gap between those two numbers. 45.4% minus 35.8% is 9.6 cents on every dollar of ad-attributed sales. On $14,164, that’s about $1,360 a month.
Be precise about what that $1,360 is, because everything below is measured against it. It’s contribution, not take-home profit — what’s left after product cost, freight and Amazon’s fees, but before storage, returns, overhead and everything else the business pays for. Real profit is smaller. I’m using contribution because it’s the number advertising is directly responsible for, and because it’s the generous version: the fees below look better against it than they would against what I actually keep.
Now put the fees against that number instead of against revenue.
| Fee, at my numbers |
Per month |
Share of the $1,360 |
| 10% of ad spend |
$507 |
37% |
| 5% of ad-attributed sales |
$708 |
52% |
| $500/mo + 5% of sales (the first quote) |
$1,208 |
89% |
| $2,000/mo base (the second quote) |
$2,000 |
147% |
| A flat software subscription |
$129 |
9.5% |
The percentage-of-spend rows are illustrations, not a claim about what anyone charges — run your own fee through the same test. But notice where they all land. Every percentage model takes somewhere between a third and half of what my advertising contributes, before any monthly fee is added on top. The first quote takes about nine-tenths of it. The second takes more than all of it, which would mean paying for the privilege of running ads.
And the simplifications run in the agency’s favour, not mine. Ad-attributed sales are generous — Sponsored Brands and Sponsored Display claim sales that other campaigns helped make. Between that and the costs contribution leaves out, the real gap is smaller than $1,360, which makes every percentage in that table worse, not better.
The mechanism, not my numbers
Your numbers will be different. The mechanism is what transfers:
What advertising actually contributes is the distance between your break-even ACoS and your actual ACoS. A percentage fee is charged on the whole sale — or on the whole spend — not on that distance.
A 5% fee costs you five points out of however many points of gap you have. I have 9.6, so it takes about half of that contribution. Run five points under break-even and 5% takes all of it. Sit at break-even — which plenty of accounts do without knowing it — and the fee is charged against zero.
That is why “is 5% reasonable?” can’t be answered in the abstract. It is reasonable or ruinous depending on a number most sellers have never calculated.
A percentage of ad spend has its own quirk: it rises when your spend rises, including when the extra spend didn’t sell anything. A percentage of sales at least moves with sales. Neither one moves with profit, and profit is the only one of the three you actually keep.
Ad sales and profitable ad sales are different things
This is the incentive question I actually care about, and it isn’t about anyone’s honesty.
Anybody can run up your ad sales. Bid higher, go broader, buy more clicks. Sales go up. At a 100% ACoS your sales can double while your bank account does nothing at all.
A fee tied to ad-attributed sales pays the same on both kinds. It can’t tell the $14,164 I have from a $20,000 that cost $19,000 to get.
I’m not saying agencies chase that. Most are trying to do good work. But if you’re going to pay a percentage of something, pay it on a number you’d be happy to see maximised.
What you’re actually buying
At this volume the work can’t be done by hand. Nobody adjusts several hundred bids a month manually, reads every search term, and watches placement performance across dozens of campaigns. So whoever manages the account — me, you, or someone you hire — is running software to do it. That’s not a criticism. It’s the only sane way to handle the volume, and I’d be wary of anyone claiming otherwise.
Which means the question was never software versus people. It’s what the judgment on top of the software is worth, and that’s a real question — judgment is the part I care most about, it’s the reason I review recommendations instead of auto-applying them, and a good operator is worth paying for. I built a judgment layer and I still don’t think it removes the need for a person.
So the honest way to put it is as a question I had to answer for myself, not a verdict on anyone’s work: to sign the first quote, I’d have needed to believe that judgment was worth 89% of what my advertising contributes. I didn’t. At a different size, with a different gap, I might have.
When an agency does make sense
The honest line isn’t “never hire an agency.” It’s that the economics stop working in the middle of the road.
If you’re running serious ad spend and moving into the more advanced side of advertising — DSP, multi-channel, real creative work, brand campaigns that need someone whose whole job it is — an agency earns its fee. At that scale a percentage is a smaller share of a much larger gap, and there’s genuine specialist work no rules engine does.
If you’re a mid-sized seller running Sponsored Products and some Sponsored Brands, that isn’t the work. That work is repetitive, and repetition is what software is for.
What I actually run
I’ll be straight about my own account, because the honest version is more useful than the brochure version.
I don’t run everything the software can do. My rule-based automations are paused at the moment — not because they stopped working, but because I’m running the AI layer on my own account to see how well it holds up. We’re always testing. I need to know all the time that we’re doing a good job for the people paying for this, and my account is where I find that out first.
Over the last 30 days that layer produced 1,509 recommendations. 692 were applied. I rejected 66. Since April it has produced just under 6,000. The rejecting is the part I’d never give up — it’s where I catch the recommendation that’s arithmetically correct and strategically wrong, like negating a term that is my category.
That’s the shape of it: software does the volume, I do the judgment, and it costs a flat monthly fee that doesn’t move when my spend does.
So when do you fire them?
Not when the reporting annoys you. When the economics stop working.
Calculate your break-even ACoS. Find your actual ACoS. The gap between them, times your ad-attributed sales, is what advertising is contributing to your business. Then put the fee next to that number.
If the fee is a small share of the gap, you’re buying a service. If it’s most of the gap, you’re not buying management — you’re buying it with your margin. Margins are margins, and 5% is a big chunk of a small gap.