Notes on building the machine.

A funnel is a sequence of small agreements. Each step asks the reader to accept one more thing: that this problem is theirs, that it is expensive, that it is solvable, that it is solvable by you, and that now is the moment. A page can present those agreements well or badly, but it cannot invent one that the offer does not contain.

So when a funnel underperforms, the useful question is not "what should the headline say". It is "which agreement is missing". Usually it is the third or the fourth: the reader believes they have the problem and believes it costs them money, but nothing in the offer explains why this particular structure solves it, or why you are the one who can install it.

That gap is expensive because it is invisible. Traffic still arrives, the page still gets read, calls still get booked, and the closing rate quietly stays where it was. The team responds by increasing volume, which multiplies the leak instead of sealing it.

The repair is upstream. Narrow who the offer is for until the promise can be specific. Name the mechanism, not just the outcome, because a mechanism is what makes a promise credible. Price at a level that lets you deliver the mechanism properly. Then decide what the buyer is actually agreeing to at each step, and write the funnel as those agreements in order.

Done that way, the copy gets shorter and the sales conversation gets shorter with it. Not because the writing improved, but because there is less left to argue about by the time someone reaches the call.

Ask a founder how sales went last month and you often get a feeling: strong, slow, weird. That answer is not laziness. It is what happens when the record of what occurred lives in four places, none of which agree, and reconstructing it costs an afternoon nobody has.

The minimum viable sales infrastructure is smaller than people expect. One place where every lead exists. A short, honest set of stages that describe what actually happens rather than what should happen. One owner per stage. One agreed definition of what moves a deal forward. A follow-up sequence that runs whether or not anyone remembers it. And a single view that shows how many are at each stage right now.

The stages are where most implementations go wrong. Teams inherit a CRM template with eleven stages and then use three of them, which produces a pipeline that is technically full and practically unreadable. Better to write down what really happens in your business, in your words, even if that is five stages, and only add a stage when you can say what decision it changes.

Once the record is honest, the pattern shows itself quickly: nearly every business has one stage where deals go to die. Sometimes it is the gap between a booked call and a held call. Sometimes it is the week after the proposal. It is almost never spread evenly, which is good news, because a single narrow failure is something you can fix on purpose.

The point of the infrastructure is not tidiness. It is that on Monday morning you can look at one screen, see where the deals are stuck, and do something about it that same day.

A service business usually grows to its first ceiling on the founder’s judgement. That judgement is real and valuable, and it is also undocumented, which means it cannot be delegated, reviewed, improved or repeated by anyone else. Every new hire has to reconstruct it by osmosis, and every reconstruction is slightly different.

The instinct is to hire a senior person who will "own" the area. That helps for a while, and then the same thing happens one level up: now two heads hold undocumented judgement instead of one, and they disagree in ways nobody can adjudicate because there is no written standard to compare against.

An SOP is not bureaucracy. It is the smallest possible answer to a specific question: what does good look like here, and how would I know. Written well it fits on a page: the trigger, the steps, the decision points with their criteria, the output, and one worked example of the output done properly. The example does most of the work.

Start with the processes that are repeated weekly and cost money when they are done badly: onboarding a new client, qualifying a lead, handing a closed deal to delivery, producing the weekly report. Write the first version by recording someone doing it and transcribing what they actually did, not what they think the process is. Then let the person who runs it edit it. Ownership of the document is what keeps it alive.

The test is simple and unforgiving. Hand it to someone competent who has never done the task and watch. Wherever they hesitate, the document is missing a decision criterion, and that hesitation is exactly the place the founder was still needed.