Step 1: Record a walkthrough of your onboarding

Open your product in an incognito window so you see exactly what a brand-new user sees. Don't be logged in. Don't have any saved state.

Start a screen recording (Loom is easiest, but any screen recorder works). Hit record, then:

1. Go to your marketing site the way a new user would
2. Click "Sign up" or whatever starts the flow
3. Walk through every step until you reach a moment where the product feels useful
4. Use the product for a few minutes after that
5. Stop recording

Narrate as you go. Out loud, say what you're doing, what you expected, what you're confused by, what you're looking for. Don't rehearse. Messy is good. The transcript is part of the audit.

If you use Loom, the transcript generates automatically. You can copy it from the "Transcript" tab on the video page.

Step 2: Pull screenshots from your recording

Scrub through your recording and grab a screenshot of every distinct screen your user passes through.

Specifically:

1. The signup page. The page with the actual signup form, after clicking "Sign up" from your marketing site. Not the marketing site itself.
2. Every setup or onboarding screen. If there's a wizard, each step. If there's a welcome modal, the modal. If there's a tour, each step of the tour. If there's a form, the form. Don't skip anything that required you to click, read, or decide.
3. The first screen where the product feels "ready." Usually a dashboard, workspace, or main app view, after setup is complete.
4. Every screen you interacted with while trying to reach a real result. Click around naturally. Try to use the product the way it's meant to be used. Screenshot every distinct page or major modal you visit.
5. The moment of value, if there's a success state. If your product shows "Congrats!" or "Your first [thing] is ready," capture that exact moment.
6. The screen right after that moment. What does the product point you toward next?
7. Any paywall, upgrade prompt, or locked feature you hit along the way. Optional, but include it if it shows up.

Most products produce 8 to 15 screenshots. That's the right range. If you have fewer than 6, you probably skipped something.

Step 3: Fill in the top three sections of the prompt

Copy everything inside the code block below. You'll fill in three sections at the top, then leave the rest alone.

What to fill in:

• Your product description. Copy the headline and subhead from the top of your marketing site. That's usually the best two-sentence description of what your product does and who it's for.
• Your narration transcript. Paste the auto-generated transcript from Loom (or your screen recorder). Write "Not available" if you don't have one.
• Anything else Claude should know. Optional. Useful for context like "we're pre-PMF," "our target users are non-technical marketers," or "we know our signup flow has a problem, focus the audit on what happens after."

Everything below the line that says DO NOT CHANGE ANYTHING BELOW THIS LINE stays exactly as written.

Step 4: Run the audit in Claude

Open a new Claude conversation. Paste the filled-in prompt. Attach all your screenshots to the same message. Submit.

Claude will produce the full audit in a couple of minutes.

The prompt

The 11 rules (reference)

If you want to learn the framework without running the audit yourself, here are the 11 rules with a short explanation of each.

Define a Same-Day Win

Rule 1: Your product must have a defined first-session win. A Same-Day Win is the smallest meaningful result that makes a brand-new user think "this was worth my time." Without one, your users are walking into a room with no destination.

Rule 2: The first win must be reachable alone. One person, one sitting, no dependencies. Every integration, data import, or "wait for your teammate" is a place users drop off.

Rule 3: The win must connect to the user's actual job. Wins framed around features ("Dashboard created") are demos. Wins framed around outcomes ("You just found where your team is losing time") are victories.

Rule 4: The win must create a reason to return. If a user wouldn't come back tomorrow after their first win, the win is too small, too abstract, or disconnected from their real work.

Clear the Path

Rule 5: The path must be three steps or fewer. Every screen between signup and the first win is a drop-off point. Cut, defer, or automate anything that isn't essential.

Rule 6: The main action must be obvious in two seconds. One dominant CTA per screen. Competing actions of equal weight force the user to decide, and deciding is where they leave.

Rule 7: The product must do the hard work first. Sample data, defaults, and pre-configuration beat blank-canvas setup every time. Show value before asking for effort.

Rule 8: The product must teach itself. If users need docs, chat, or a demo call to get through onboarding, the product is offloading its job to humans.

Prove It Worked

Rule 9: The product must acknowledge the win. A win the user doesn't notice is a win that didn't happen. Name the outcome, in the user's language, at the moment it occurs.

Rule 10: There must be a clear next step after the win. Leaving a user who just won to wander a dashboard is the fastest way to lose them. Design the next pull.

Rule 11: Users must be able to share their results. The product's output should do your marketing. The moment sharing is effortless, growth becomes a byproduct of value.

How the score works

Your audit produces four scores:

• The overall score, a weighted average out of 100
• Three sub-scores, one per section, each out of 100

The weighting reflects what matters most. Defining a Same-Day Win is the foundation everything else sits on, so it's worth 40% of your overall score. Clearing the path is worth 35%. Proving it worked is worth 25%, because it's the easiest to fix once the other two are in place.

Your tier tells you what kind of problem you're looking at:

• Solid foundation (80 to 100): Your activation works. You're refining, not rebuilding.
• Clear gaps (50 to 79): One or two things are dragging the whole system down.
• Compounding issues (below 50): The problems are connected. Fixing one in isolation won't move the needle.

Your weakest link is the section with the lowest sub-score. That's where most of your leverage lives.