Billings are not revenue

Before anything else can be argued honestly, one number has to be thrown away.

A local service client pays $6,500 a month: a $3,500 retainer plus $3,000 of ad spend. If that spend runs through the agency’s card, it lands in the agency’s bank account, and a great many agency owners quietly count it. It makes the business look twice its size on a slide.

It is not income. It arrives and leaves the same week, on its way to Google and Meta, and the agency is a courier for it. The industry term for what is actually left is agency gross income — revenue minus everything that passes straight through to a third party. Parakeeto puts it plainly: an agency doing $5m in revenue with 50% pass-through has $2.5m to actually run on, and every budget set against the bigger number is set wrong.

What actually happens to $3,500

Nine lines stand between the money arriving and the owner keeping any of it. Here they all are.

The example below is a real shape: one local service client on a $3,500 monthly retainer — roughly the floor at which this work can be done properly, and close to the minimum we argue for elsewhere. Delivery runs about fifteen hours a month across a specialist and an account manager.

Fourteen point six percent is not a disaster. It sits inside the band that Move at Pace puts on well-managed agencies for 2026, and near the low end of the 15–20% TMetric reports for generalist shops. It is a working business. It is also nothing like what the person doing the delivery work imagines is happening.

The specialist sees $525 land in their account and a client paying $3,500. The arithmetic they do in their head is that somebody upstairs kept $2,975. The real number is $510, and it is the only line on the page that can go negative.

The middle nobody logs

Delivery work is visible by design. Almost everything else is invisible by design. That asymmetry is the whole misunderstanding.

If your job is to run the ads, you can see the ads. You can see the report you wrote and the call you took. What you cannot see — because it is deliberately kept away from you so you can concentrate — is the layer underneath: the senior review that caught the mistake before the client did, the two hours of access wrangling, the invoice that had to be chased twice, the insurance premium, the software renewal, the client who left owing a month.

None of it is glamorous and none of it can be skipped. Skip the QA and you lose the client. Skip the ops and the access breaks. Skip the sales and the book shrinks by 20–30% a year through ordinary churn. The reason it feels like it does not exist is that a well-run agency spends real money making sure the delivery team never has to think about it.

There is a number that puts a floor under this. TMetric’s 2026 benchmark study of 250+ firms found that manual time capture reaches only 67–68% of billable hours, and that 23% of tracked hours never make it onto an invoice at all. That is not laziness. That is what happens in a business where the work moves faster than the paperwork — and the gap is paid for out of the same $3,500.

What good actually looks like

If you only ever see your own P&L, you cannot tell whether 14% is a triumph or a warning. Here is the outside view.

Three ways an agency goes broke while looking busy

One — the wrong denominator

You count the ad spend. Revenue looks like $6,500 a client, margin looks like 8%, and you conclude you need more clients. You do not need more clients. You need to stop counting money that was never yours. Everything downstream of this error is unfixable until the error is fixed.

Two — paying for the person, not the role

This is the expensive one, and it almost never starts as a mistake. Someone talented arrives, you want to keep them, and you offer a share of revenue that reflects your hopes for them rather than the roles they will actually perform. A 70% operator share sounds generous and survivable. Run it through the ledger:

The tell is that a high share is defensible only when the person is also carrying the roles it is paying for. Someone who sources their own clients, runs them, and trains their replacement has earned most of the card. Someone doing execution on clients you found, with tools you bought, using a method you wrote, has earned one slice of it — and being excellent at that slice does not turn it into three.

Three — a price built from delivery hours only

You estimate fifteen hours, apply a multiple, and quote. The middle was never in the number. Then QA, ops, the unbilled favour and the churn replacement all have to come out of a margin that was sized for none of them. This is why agencies can be busy, well-liked, growing, and quietly unprofitable for years: the price was wrong on the day it was set, and volume only multiplies the error.

The value-share rate card

Stop negotiating with people. Publish what each role is worth and let anyone check the arithmetic.

Every dollar of a retainer is doing one of five jobs. Write down what each job is worth once, apply it to everybody including yourself, and most compensation arguments stop being arguments and start being lookups.

Four rules, or the card will not hold

• You earn the roles you perform, not the roles you hold. Titles are not shares. A person who stops doing account management stops earning that 15% the month they stop.
• Origination follows whoever actually sourced the client. In most small agencies that is the owner, on every single account. Say so out loud — it is the honest explanation of where the owner’s money comes from.
• The mentor share is temporary by construction. It exists to make training somebody financially rational, and it ends when they no longer need it. A permanent override is a tax, not a mentorship.
• Changes are dated, noticed, and never retroactive. New terms take effect on the first of a month after written notice. Money already earned is never reopened. Break this once and nobody believes the card again.

Here is the part that surprises owners: if you delegate all five roles, your profit is zero. The house’s 45% is cost, not income. An owner’s profit is the roles the owner still personally performs — which is why an owner who stops selling stops earning.

The levels, priced

A level is a set of demonstrated behaviours. Attaching a price to each one is what stops the ladder from being a motivational poster.

We describe careers here on a nine-level ladder — the full version lives in The 9 Levels of Business Mastery. Four of those levels are the ones that touch a client retainer, and those are the four the rate card prices:

• Level 3 — specialist. Does the work. Optimises the ads, fixes the listing, ships the page. 15%.
• Level 4 — account manager. Owns the client, not just the tasks. 15%.
• Level 5 — people manager. Hires, fires, and checks other people’s work. 5% as an override while actively mentoring, on top of whatever else they do.
• Level 6 — brand and P&L owner. Knows the cost and margin of each account and puts the brand into the market. Paid out of the house share, because it is the house.
• Level 7 — agency owner. Sources the clients, carries the risk, keeps the residual. Not a share — the remainder.

The failure this prevents is specific and common: granting Level 7 economics to somebody operating at Level 3, in the hope that the money will pull them up the ladder. It does not. Money that arrives before the behaviour removes the reason to develop the behaviour, and every month it continues makes the correction more expensive and more personal. Pay the level demonstrated, publish what the next level pays, and let people climb on purpose.

What you cannot put in a contract

Everything above is an attempt to price work precisely. This section is about accepting that the attempt can only ever get you most of the way.

A rate card is a good instrument and a limited one. It can tell you what running an account is worth. It cannot tell you what it is worth when somebody stops what they are doing to answer a question they solved eighteen months ago, or records the walkthrough once so that nobody has to ask again, or tells a colleague their work is not ready yet — kindly, and in a way that makes the next version better.

None of that is in anybody’s contract here. It is not billed, not tracked, and not separately paid. I am not paid for it either, and I do more of it than anyone.

Why you screen instead of enforce

You can mandate rules. You can write laws, policies, SOPs, and consequences, and people will comply with them exactly as far as the enforcement reaches. What you cannot mandate is that somebody be helpful. You cannot mandate that they be a good person, or that they care whether the thing they are building actually works for the person on the other end.

So you do not try. You put the effort at the front door instead. You screen for character, because character is the only mechanism that produces the behaviour you cannot require — and because a person who has to be compelled to help a teammate will find the edge of any rule you write within a fortnight.

This is not softness. It is the harder and more expensive path: it means turning down capable people, and it means the hiring bar is a judgement about someone’s heart that you can be wrong about. But the alternative is an organisation where everything worth having has to be specified, and nothing that was not specified ever happens.

“So I’m working for free?”

This objection comes up, and it deserves a straight answer rather than a slogan.

No — because the investment ran in the other direction first. Before a single dollar of revenue share was calculated, somebody handed you a method that was already written down, clients you did not have to find, a stack of tools you did not have to buy, a brand that makes your call get answered, and somebody senior who reads your work before a client sees it. The first client trusted you because they trusted the person who introduced you. That is a real balance, and it was extended on credit.

Mentoring the next person is how that balance gets settled. Not because a contract says so, but because it is the only currency the debt is actually denominated in. Anyone who counts only the hours flowing out and none of what flowed in is doing exactly the arithmetic this whole article is arguing against: measuring one visible line and ignoring the eight invisible ones.

Some things are meant for loss

There is a category of thing in every good business that will never appear as a return in any period you can measure. The hour spent with somebody who then leaves. The training that gets used by a competitor. The article that helps an agency owner you will never meet and never invoice — this one, for instance.

Carrying that deliberately, as a line you accept rather than a leak you plug, is what separates a firm that compounds from one that merely operates. Everything else in this piece is arithmetic, and arithmetic is the easy half.

Run this on your own book

Six steps. None of them need a consultant, and the first four are an afternoon.

• Restate every number as agency gross income. Strip pass-through ad spend, subcontractors, and anything you merely handle. Whatever that total is, it is your business. The other number was never real.
• Price one client all the way down. Pick your most typical account and fill in all nine lines. Guess where you have to, but guess in writing so the guess can be corrected. Most owners find their profit on the page for the first time here.
• Cost an hour honestly. Fully loaded salary divided by 2,080, then divided again by your real utilization — 72%, not 100%. That is what an hour reaching a client actually costs you.
• List who performs which role on each account. Not job titles. Roles. You will find accounts where two people are being paid for the same role, and accounts where a role is not being performed at all.
• Publish the card. Internally at minimum. The moment a share is a lookup rather than a negotiation, the politics drain out of it — and you can no longer quietly give a different deal to whoever asked most persistently.
• Apply it to yourself first, and in public. Write down which roles you personally perform and what that entitles you to. An owner who exempts themselves from the card has not published a standard; they have published a rule for other people.