Isolated funnel checklist

An isolated funnel is one ad promise, one matching page promise, one primary CTA, and one measurement goal.

We wrote this after sending paid clicks at a generic homepage. The homepage has to do every job: explain the product, start chat, point at docs, and convert. A 300x250 cannot. If the banner says "run ads from Cursor" and the landing talks about six other things, you cannot tell whether the click failed or the page did.

This checklist is operator-judged. It is how we set up tests. It is not a claim that isolated funnels always convert.

The ad headline and the page H1 should repeat the same sentence, or a close rewrite of it. The primary button should complete that sentence. The secondary capture (email, prompt pack, checklist) can be quieter. It cannot introduce a second product story.

Do not rotate a broad generic creative into a specific page test unless that creative is the holdout control. Mixing "AI chat", "no sales call", "agent testing", and "API quickstart" on one page is how a readout becomes a blur.

These are the first-wave promises from the June 2026 copy plan. Pick one per campaign cell.

• ai-chat-control: Run ads from your AI chat.
• no-friction-buying: No sales call. No IO. Start the first test online.
• builder-tool-workflow: Launch a small ad test from Cursor or a coding agent.
• autonomous-agent-operator: Let AI run the ad buy.
• growth-test-matrix: Test creative, audience, and page together.
• api-quickstart-intent: A media buying API for your agent.

Name the campaign so the angle is in the first token. See the campaign naming sheet.

Isolated funnel, one cell

[ ] Audience named (who the copy is speaking to)
[ ] One job or pain (not five)
[ ] Primary promise written as one sentence
[ ] Ad headline repeats that sentence
[ ] Landing H1 repeats that sentence
[ ] Primary CTA completes the promise (signup, start test, open docs for that job)
[ ] Secondary CTA is a lower-friction capture, same story
[ ] Lead magnet matches the ad (prompt pack, checklist, not a generic newsletter)
[ ] Proof block uses product facts only (prepaid, no IO, no sales call, API/MCP)
[ ] Conversion goal chosen: email, signup started, account created, funding link started, or campaign created
[ ] Disqualifier written: what this page must not drift into
[ ] Click URL is the isolated page, not the homepage
[ ] Campaign name is angle__targeting__inventory__pageTemplate__wave
[ ] Daily budget set ($20 is the default first test)
[ ] Destination pre-check run before submit
[ ] After launch: judge real visits and the conversion goal, not click-through rate

If any line is blank, do not activate. A cheap test that mixes messages is still an expensive lesson, because you will read the wrong winner.