Director Chair
All groups

Each variant gets an id, a hypothesis, the one thing it changes, what stays fixed, a cost, an observation window and the result that would justify the next test.

Test and publish

One variable per test

A result you can read: you know why the winner won.

Use when
Testing hooks and formats, organic or paid.

Prompt instructions

Give me a step-by-step plan I can follow this week, for a short video about [your product], for [your audience]. The technique is one variable per test: a result you can read: you know why the winner won. How it works: each variant gets an id, a hypothesis, the one thing it changes, what stays fixed, a cost, an observation window and the result that would justify the next test. It is right for: testing hooks and formats, organic or paid. Ten creatives that differ in everything can find a winner and still not tell you why. Stage A explores formats. Be concrete: numbers, seconds and settings, no theory. Ask me one question first if something you need is missing.

The same prompt works in Claude, ChatGPT and Gemini.

Watch out

Related

Sources

Own words from Director Chair field notes on hooks and ideas and Director Chair notes on AI UGC army (Jake Castillo) Checked 2026-11-13.

For agents: this card as markdown, as JSON.