The landing page teardown checklist

Why the order matters

The promise decides whether anybody reads the rest. The proof decides whether they believe it. The ask decides whether they act. The mechanics decide whether the page gets a chance at all. And what happens after the click decides whether the conversion was worth having.

Run them out of order and you will spend a week on form fields for a page nobody scrolled past, which is the most common way a teardown wastes its own time. Work down the list, stop at the first pass where you find three failures, and fix those before you look further.

Before you open the page

1. Name the one number, and get today's figure. Write it as a percentage with a denominator: visitors who start the form, visitors who complete signup. A page without a named number cannot be improved, only redecorated.

2. Read what brings the traffic. The top 10 search queries, the live ad headlines, the referring pages. The page's job is to finish the sentence the visitor started somewhere else, and you cannot do that without knowing the sentence.

3. Get the device split, then do the whole teardown on the majority device. Most teardowns happen on a 27-inch monitor for a page where two thirds of the traffic is a phone on a slow connection.

Pass one: the promise

4. Can a stranger say, in five seconds, what this is, who it is for and what happens next? Those three questions are Donald Miller's grunt test, and they are the fastest check on the list. Nielsen Norman Group's 2018 eyetracking study found 57% of page viewing time happens above the fold and 74% in the first two screenfuls (NN/g, April 15, 2018), so the three answers have to be there, not further down.

5. Does the headline use the words from the ad or the query? Where possible, the exact words. A visitor who typed a phrase and does not see it within a second assumes they are in the wrong place.

6. Count the distinct claims above the fold. If there are three promises, the page has none. Pick the one that matches the traffic and move the others down.

7. Open a brand new account and time how long until the page's promise comes true. If the page promises a report and the product needs a developer and two weeks before it produces one, the leak is not on the page. It is in the gap, and the page is writing checks activation has to cash.

8. Screenshot the first screenful on a real phone with the cookie banner, the chat bubble and the announcement bar all live. That is the page your visitors see. It is often not the page anybody on the team has looked at.

9. Is the alternative named, including doing nothing? Most buyers are not choosing between you and a competitor. They are choosing between you and a spreadsheet they already have. A page that never mentions the spreadsheet is arguing with nobody.

Pass two: the proof

10. Is each piece of proof next to the claim it supports? A testimonial carousel at the foot of the page proves nothing, because nobody reaches it and nobody knows which claim it answers.

11. Does every number on the page have a source and a date? If it does not, remove the number. One unsourced figure makes a reader suspicious of the sourced ones.

12. Is there one customer on the page who looks like the reader? Same rough size, same sector, ideally the same tool they are already paying for. One close match beats eight famous names.

13. Is any of the actual work visible? A real screenshot, a sample output, an anonymized report. Showing the thing beats describing it, and it is the cheapest proof you own.

14. Are the logos earned? A logo with no sentence next to it is decoration. Keep the ones a reader would recognize as peers and cut the rest, and never put a logo on the page for a company that is not a customer.

Pass three: the ask

15. Count every clickable thing above the fold. Navigation, the logo, chat, secondary buttons. Each one is an exit. On a paid landing page, most of them should not be there.

16. Count the form fields, then justify each one. The test: does somebody act on this answer within a week? If the answer sits unread in the CRM, it belongs in enrichment or in the second session. For scale, Baymard's 2024 benchmark found ecommerce checkouts averaging 11.3 form fields across 5.1 steps where most sites need only 8 fields in total (Baymard Institute, June 26, 2024), and B2B signup forms are usually worse.

17. Does the button say what the visitor gets? "See your own data" rather than "Submit". The label is the last thing read before the decision.

18. Is the form on the page, or behind a click? On the page, unless you have tested otherwise for this specific traffic. Every extra step before the form loses the people who were about to fill it in.

19. Is there exactly one smaller ask for people who are not ready? One. A page with a demo button, a trial button, a newsletter box and a download is a page that has decided nothing.

20. Is there a price, a range, or an honest sentence about why there isn't one? Silence on price sends the reader to the competitor who published theirs, and they do not come back.

Pass four: the mechanics

21. Measure Core Web Vitals at the 75th percentile of real users. Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, Cumulative Layout Shift under 0.1 (web.dev, updated October 31, 2024). Deloitte and Google found a 0.1 second improvement in mobile speed lifted lead generation form submissions by 21.6% across 37 brand sites (web.dev, June 24, 2020), which makes this the best-paying hour on the list.

22. Test on a mid-range Android on a throttled connection. Your laptop on office wifi is not a measurement, it is a compliment.

23. Does the page still make sense with images blocked? The promise should be in text. If the headline is baked into a hero image, a slow load shows a blank screen with a button.

24. Check the input details on a phone. Numeric keyboard for numbers, email keyboard for email, tap targets at least 44 pixels, real labels rather than placeholder text that vanishes when typing starts.

25. Is form start tracked separately from form submit? Without that one event you cannot tell a copy problem from a form problem, and you will argue about which it is for a month.

Pass five: after the click

26. What happens in the 10 minutes after someone submits? Name the person or the workflow that sees it, and what they do. A page that converts into a queue nobody watches has not converted.

27. Does the thank-you page do any work? The next step, how long it will take, and one thing worth reading while they wait. It is the most attentive screen on the whole site and usually the emptiest.

28. Does the first email arrive within a minute, and does it name one next action? Not a brochure. One link to the one thing that makes the product work for them.

29. Is an exit survey running, and is somebody reading it weekly? One question: "What were you looking for that you didn't find?" Twenty answers in the visitors' own words will set your next month's work better than any checklist, including this one.