For founders
What a free teardown looks like
A real-format example on a made-up AI app: five onboarding questions, a sign-up wall and no sign of what you get. Three fixes, one page, with the activation benchmark that explains why they matter. This is what lands in your inbox.
Prashant Kumar Singh · · 4 min read
In short
- A teardown is one page: what to fix, what to cut, what to build first, with a scope and price at the end.
- The example app asks five questions behind a sign-up wall before showing anything; the three fixes are show a result first, ask one question not five, and let people try before the email wall.
- Lenny's Newsletter survey puts average SaaS activation around 36%; showing value before sign-up is the most common lever to move it.
People ask what they would get from a free teardown before sending anything, which is fair. So here is one, in its real format, on an app I made up so nothing in it is confidential. Call it Draftly: an AI tool that writes a first draft of a cold email from a LinkedIn profile URL.
Before
The landing page has a headline, a screenshot and a "Get started" button. The button opens a sign-up form: email, password, name, company, role. After sign-up, an onboarding flow asks five questions: industry, target persona, tone, email length, and how the user heard about Draftly. Only then does the product show the box where a URL can be pasted. Nowhere along that path has the user seen a single draft email.
Seven screens between arriving and seeing what the product does. Every one of them is a place to leave.
Fix 1: show a result first
Put one worked example above the fold: a real-looking profile on the left, the draft email Draftly wrote on the right. Users decide in seconds whether the output is good enough to want, and the example answers that before anything is asked of them. This is the single highest-leverage change for an AI product, because the output is the product.
Fix 2: one question, not five
Of the five onboarding questions, one affects the first draft: the target persona. Tone and length can be defaults the user adjusts after seeing a draft. Industry can be inferred from the profile pasted. "How did you hear about us" is for Draftly, not the user, and belongs in a later email. One question, then the product.
Fix 3: let people try before the wall
Let a visitor paste one URL and get one draft without an account. Ask for the email to save the draft or write a second one. The user who has just read a draft written for their prospect will give an email. The user staring at a password field has not seen a reason to.
After, and the number behind it
Lenny's Newsletter surveyed SaaS companies on activation and found an average around 36%, with wide variation by category. The most common way products move that number is by shortening the distance between arrival and the first moment of value. Draftly's fixes cut the path from seven screens to two: see an example, paste a URL. Everything else moves to after the first draft, where a user who has seen value will answer.
What to send for yours
A link, a prototype or a one-page idea. Say who it is for in one sentence and name the one thing you are unsure about. The teardown comes back within a few days, in this format, and it is yours whether or not we work together.
Questions people ask
- What is included in a product teardown?
- One page: what to fix, what to cut, what to build first, and a scope with a fixed price for the first build. It is written from your product, prototype or idea and is yours to keep.
- What is a good activation rate for SaaS?
- Lenny's Newsletter's survey puts the average around 36%, varying widely by category. The usual lever to improve it is shortening the path from arrival to the first moment of value.
- Should an AI product show results before sign-up?
- Yes. The output is the product, so one worked example above the fold, and ideally one free use before the email wall, lets users decide they want it before they are asked for anything.
If you're a founder
Send your onboarding. The teardown will come back in this format, with your three fixes instead of Draftly's.
If you run an AI lab or data team
The task-review equivalent is the same shape: what fails, what to cut, what to change first, in writing within 48 hours.
Sources