Three weeks on a dead thesis

On August 10 an agency owner told me, kindly, that the problem I was solving was not a problem she had. I wrote it down. Then I spent the next three weeks building a dial list, an email campaign, a practice-management feature and a demo flow, all aimed at her.

This is not a story about denial. I knew the call had gone badly. The problem was structural: every artifact in the repo (the ICP document, the outreach scripts, the metrics, the kill gate itself) was written for the old thesis, and the cheapest thing to do each morning was the next item on the old list. The project had momentum and the momentum pointed at a wall.

Three things would have stopped it sooner.

A kill gate that fires on one call, not ten. Ten was the number I had picked for a statistical reason. But the first call was not a data point, it was a domain expert explaining the domain. Gates should have a fast path for “the premise is wrong”, separate from “the numbers are bad”.

A weekly question with a forced answer: who pays, and why. Not “who has the pain”. Every week I could answer the pain question and could not answer the payment question, and nothing in my process made me write the second answer down.

Fewer artifacts. Each script, list and feature was a small sunk cost that made the next one cheaper than starting over. The agents made this worse, not better: when building is nearly free, building instead of asking is the default.

The app went to Apple for review on September 2. I set the project down on September 3, with the real gate at one conversation of ten. The next project has a gate of three named answers out of ten asks, and the question about payment is inside the ask.