What breaks first when an agency grows

Three things break, in a fixed order. Delivery quality goes first. Cash follows it. The founder goes last, and by the time that happens the other two have been broken for months. The order holds because each break is caused by whatever made the previous stage work.

What breaks first when an agency grows

Quality, and earlier than anyone expects.

While one person does the work, the standard is taste. You know what you would ship because you are the one shipping it, and that knowledge never has to leave your head to get used.

The second person changes that. The standard now has to travel out of your head, through a brief, into someone who has never watched you reject your own first attempt. What comes back is usually on brief and still wrong, and you can only say why once you are looking at it.

So you fix it yourself. The first few times that is the right call. Then review starts taking longer than the edit took, work stacks up behind you, and the editor you hired to take work off your plate is waiting on you to clear it.

Write the standard as a rejection log

Style guides describe the ideal. They're pleasant to write and nobody uses them, because an editor working late does not need a paragraph about brand feeling. They need to know whether this cut passes. (Ask an agency to show you theirs and watch how long it takes them to find the link.)

A rejection log works better. Every time you fix something yourself, that fix is an unwritten rule you have just paid for. Write it down in the form it happened in: what you received, what was wrong with it, what it should have been, with a timestamp on the file. Send the list to the person whose work you changed. Nobody enjoys doing this. A short run of it beats anything you would have written up front, because every line was earned.

Then separate the checkable part from the preference. A brief can carry the checkable part without you standing over it: safe area for captions, where the hook has to land, deliverable naming, audio levels, aspect ratio. Those are pass or fail, and someone who isn't you can enforce them. Preference is the part that needs your eyes, and shrinking it is the whole job.

One more, and it is the one people resist. Review earlier. The cost of a rejection scales with how finished the work was when you rejected it, so asking for the sequence before the polish is a small interruption that saves a rebuild. It is also the only version of quality control that survives volume. The hours move to the point where they are cheap.

Delivering one thing to a standard, repeatably, was always the real business underneath the acronym. I wrote about why the category got a bad name while that part stayed true.

The cash gap opens exactly when you win

Retainers pay in arrears. You deliver a month of work, invoice at the end of it, then wait for however long the client's accounts process takes. Editors expect to be paid on time, and the good ones have other options, so the people who leave first when payment slips are the people you least wanted to lose.

That gap between money leaving and money arriving is manageable at one client. It widens with every new one, because a new client means editor cost before the first invoice has even been raised. Which produces the outcome that catches people out: signing two clients in a month can leave your balance lower at the end of it than at the start. You can be profitable on paper and short of money in the same week, and selling harder widens the gap instead of closing it.

Four things change the shape of it.

• Bill the first month up front, or split it. It is a payment term, and holding it in the first conversation takes the same nerve as quoting a rate without hedging.
• Stagger due dates so payouts do not all land in one week. Costs nothing, and it moves the worst day of your month.
• Size the buffer in editor payouts. How many could you cover with nothing coming in? That number decides whether you can take the next client.
• A late-paying client is a cash problem. Owners who file it under relationships sit on the reminder for weeks.

Tax registration belongs on the same list. Most tax regimes have a turnover line where invoicing as a sole proprietor stops being enough, and some require registration from the first sale, and crossing it brings an obligation with a date attached and a compliance cost that does not care how busy the month is. Knowing which side of that line you are heading towards is a cash decision months before it becomes a paperwork one.

The paper is weaker than you think

Worth saying if you are under eighteen and signing things.

A minor's agreement is weak protection almost anywhere, and the part people miss is what that does to the paper you sent out. Depending on where you are it may bind neither side, or only the adult across the table. So the editor NDA may be worth less to you than the reassurance it gave you when you sent it.

I am not a lawyer, this is not advice for your situation, and the rules differ depending on where you and the client sit. The operating conclusion still holds: put the protection into how you work. Release footage in tranches instead of handing over the archive, and pay against delivery milestones. Client contact should sit in channels you control, which does more for you on the day it matters than a non-solicit clause you could never afford to argue about.

The founder goes last, and quietly

Attention is the only input you cannot get more of. You can hire another editor, and you can go and find another client. Your attention has a ceiling that does not move, and every system that routes a decision through you is spending it.

The failure doesn't look like exhaustion at first. It looks like this: the urgent work gets done every single day and the important work goes untouched for a month. Review queues are urgent, because a person is waiting at the end of one. Sales is important, and nobody is waiting. The queue wins, and it wins in a way that feels like working hard. By the time it shows up as tiredness you have already lost the month.

The repair is routing, and it is dull. Decide what genuinely cannot leave your desk. For me that is the standard itself, pricing, and the decision to end a client relationship. Everything else gets an owner or a default.

A default is a written answer to a question that would otherwise arrive in your DMs. If a client asks for a fourth revision, here is what happens. If a lead asks for a discount, here is the answer, and it is the same answer every time. You decide it once, and then it stops being a decision. Write them down as they come up and the DMs get quieter.

Giving one person authority to approve inside a band is the other half. If a cut passes the checkable list and the client signed off on the previous three, it ships without you. You'll be wrong sometimes. A few weak cuts going out cost less than a queue that only clears when you sit down at it.

Why all three are survivable

None of them is a talent problem, although each one feels like one while it is happening.

Quality breaking is a documentation problem, and cash breaking is a timing problem. Both of those fit on a page. The attention one is a routing problem, which is harder, because the routing runs through the person who would have to decide to fix it. What all three share is that they are visible well before they are fatal, if you know what to look at.

The reason they get read as character failures is that they arrive during a good month. A quiet month hides all three, because none of them is under any load. I have been building since fourteen, and the pattern I trust most is that the trouble lands in the month straight after a strong sales month.

What to do this week

Two things, both boring.

Open the last five deliverables you fixed yourself. Write down, for each one, what you changed and what you should have received instead. That list is your standard, in the only form anyone will read.

Then open a calendar and mark every date money leaves this month and every date money is due to arrive. Find the widest gap between them. That number is what decides whether you can take the next client.