Four times in the chair. Eight of the stories written so far.

A support metric that was the ceiling on the company

Churn sat on a dashboard for about a year, owned by the support team. Nobody had multiplied it out. Four and a half per cent a month compounds into losing forty-three of every hundred customers in a year, which is not a support problem. It is the maximum size the company can ever be.

The activation metric was backwards

Every onboarding in the industry tells you to get new users creating projects. The model said each project a trial user created cut their odds of converting by eighty per cent, and each update to an existing one raised them. Creating project after project is not using a tool. It is shopping.

We put it on a stage against the experts

A conversion specialist is a salary most small companies will never afford. So we built the specialist instead, and then ran it live against human specialists at a conference to make anyone believe it. Beating them was never the point. They were never the customers.

The discount was not a pricing decision. It was a deferred one.

A bootstrapped services business discounting to close a net-ninety cash gap, which was a correct answer to the constraint she could see.

What was not in the model was what the discount was buying: dunning, and the overages when a client arrives with data nobody checked. One dirty dataset does not delay one project. It forms a queue.

We called it a tamper-proof wallet

Nobody had decided who owns a medical record, so everyone treated it as a technical problem. In the US, if your employer chooses your insurer, your history is attached to your job. The hard part was not the encryption. It was that in 2020 nobody knew what encryption meant.

The one that stopped

Augmented reality on phones, four years before the platform existed, with a dated deck to prove the call. It did not fail on the idea or the technology. It failed on a founding team that never committed, and I had done no work at all on whether the three of us wanted the same thing.

I could see what it would cost, and I signed it anyway

Three reductions had already happened and engineering had been protected through all of them. This one was for cost.

My read was that cutting there would leave the company depending on a handful of individuals to a crippling degree. Without it there was no capital, and without capital none of the other decisions existed. The only page here where nothing was misnamed and the choice was still hard.

The only critic was paid out of the tuition

The studios gave me two complaints in 2007 and both sound like curriculum. They were not.

A student made something, one person saw it, and that person's salary came out of their fees. Nobody in the building could say no and mean it. Also the page where I show you how much of my own best number is arithmetic rather than teaching.

Thinking out loud

One unbroken chain of association, recorded in a single sitting, on a subject nobody has paid me to have an opinion about. Six jumps, three dead ends left in, and it lands on a question rather than an answer. Every other page here shows you a conclusion.

What it was called, and what it was

All six pairs on one page, so you can see whether the pattern is real or whether I have told you three good stories. Nothing on it is measured, and it says so.

A time I was wrong about the problem and the client was right

Every failure on this page is a failure of persuasion, of naming, or of a founding team that never committed.

None of them is a case where I read the situation wrong and somebody else read it correctly. Nineteen years in, that cannot be true, and a record with no instance of it is a record you should discount.

It is not written because the story I have in mind is not mine to tell without asking. I am asking.