The Wall

Getting to three million is hard in a way that makes sense. You win clients, you deliver, you hire, you grind. It is exhausting and it works.

Somewhere after that, the effort stops converting. Headcount goes up and output doesn't follow. You add a client and something unrelated breaks. The week fills with decisions that all seem to need you, and none of them are the ones you meant to spend the week on.

The instinct is to read this as a capacity problem, because that is what it looks like from inside a busy week.

Capacity problems get worse gradually as you grow. This one arrives at a specific size and does not respond to hiring, which is the tell that it is a different kind of problem.


Every Operation Has an Integration Layer

Between what sales promised and what the floor does, something has to translate.

Somebody knows that this client's orders need a second look before they ship, that this account's cutoff is really an hour earlier than the contract says, that when those two things collide you call the client rather than shorting the order. Somebody holds the exceptions, routes the judgment calls, and remembers why each one exists.

In a young operation that somebody is usually the founder, and it works extremely well. A person is a fast, flexible, infinitely configurable integration layer. No implementation project required.

That translation function is a system. It is just a system that happens to be implemented in one human being, which is invisible for as long as the human being keeps up.


The Part You Cannot Hire

You can hire pickers. You can hire a supervisor, a customer service lead, even a capable operations manager.

What you cannot hire is somebody who was in every client conversation since inception, because that is not a role, it is an accumulated history. So the new operations manager escalates, which is correct, and the escalations arrive at the same desk they always did.

The wall is the throughput of that desk. It sits at whatever size the accumulated exceptions exceed one person's working memory, which for most operations lands somewhere in the low millions, and no amount of hiring underneath it moves it.

Everyone experiences this as the founder being a bottleneck, which frames it as a delegation problem, and delegation cannot transfer knowledge that was never written down.


It Gets Worse the Day You Buy Software

This is the part I watched from the implementation side, repeatedly.

An operation decides the answer is a real system, which it is. Configuration starts, and the system needs the rules: what happens when an order is short, who approves an exception, which clients get which treatment and why.

Those rules exist. They have run the business for years. They have simply never existed anywhere except in the head of the person who is also, right now, still running the floor.

So the implementation queues behind one calendar. The project stalls, and it looks like a resourcing problem or a vendor problem, and it is neither.

The operation is being asked to write down the thing it built specifically so it would never have to write it down.


Not Only Founders

The same structure shows up without a founder anywhere near it.

In a lot of distributors it is one operations veteran with twenty years in the building. They know which retailer's routing guide is enforced and which is theater, which supplier's traceability paperwork actually gets audited, which customer's SLA has teeth. Nobody wrote any of it down because nobody needed to.

That business has the same ceiling and a worse version of the risk, because the person holding it is closer to retirement than to founding another company.

An operation that runs on one person's judgment has an expiry date attached to a human being, whether or not that person owns the place.


Getting past this has never once involved cloning the founder. It involves writing the exceptions down, discovering the list is both longer and more contradictory than anyone expected, and resolving it into something a system can hold.

That work is unpleasant and it is the actual price of the next tier of growth. It is also, not coincidentally, the same document a WMS evaluation needs and almost never gets.