SaaS landing page copy: three annotated teardowns

The hardest version of a landing page to write is the first one, because you are writing for visitors you have never met, with no traffic to test on, and nothing honest to put in the testimonial slot. Almost every guide to this subject assumes the opposite situation: a page with enough traffic to split-test, a designer, and a wall of customer logos. This one assumes what you actually have, which is a page, a rough idea of who it is for, and maybe forty visitors a month.

So this is a teardown rather than a lecture. Three pages taken apart section by section, with the copy that was there, the copy that replaced it, and the reasoning in the margin. The pages below are composites. No real page is reproduced word for word, the products are invented, and the patterns are the ones that keep turning up in the landing pages founders put through our free landing page roast.

What is the only job of your hero?

The hero has exactly one job: within five seconds, a stranger should be able to say what the product does, who it is for, and what happens if they click. Everything else up there is decoration, and decoration is what you cut first.

That five-second budget is not a rhetorical flourish. Nielsen Norman Group's reading research puts the share of words an average visitor gets through at somewhere between 20 and 28 percent, and their later eye-tracking work on scrolling found about 57 percent of viewing time spent above the fold, and 74 percent inside the first two screenfuls. People do scroll now. They just spend most of their attention deciding whether scrolling is worth it, and that decision is made on your headline, your subhead, and your button label.

The most common way a founder hero fails is not ugliness or a weak offer. It is category poetry: sentences built on reimagined, effortless, all-in-one platform, or built for modern teams. Those are verdicts about your own work, delivered by you, to someone with no reason to accept them. Worse, they are interchangeable. Swap the logo and the same sentence sells a CRM, a code editor, or a mattress.

Which headline formulas survive real visitors?

Five shapes hold up under a stranger's five seconds, and each one fails in a specific, predictable way. Pick the shape that matches what is genuinely true about your product rather than the one that sounds most impressive, because the failure column is where founders end up when they choose on sound.

One test settles most headline arguments. Read the sentence and ask whether a competitor could put it on their own page without lying. If they could, it is not a headline, it is a category label. The fix is almost always to add the specific noun you were avoiding because you worried it made the product sound small. Postgres instead of databases. Freelance designers instead of teams. Small does not lose you readers. Vague does.

The words themselves should not be invented at your desk. The most reliable headline I have written came out of a sentence somebody said on a call, almost verbatim, complete with the slightly awkward phrasing that made it sound like a person. If you have not run those calls yet, the question set in customer discovery interview questions is where the raw material comes from.

How should you order the middle of the page?

The middle of the page has one structural rule: answer objections in the order visitors think of them, not in the order your features happen to be listed in your codebase. A feature grid is what you build when you have not decided what the reader is worried about, and readers can tell.

Six questions run through a visitor's head, roughly in this sequence. Each one should own a section, and a section that answers none of them is a section to delete.

1. 01Is this for someone like me?

   The hero, by naming the audience and the job in one sentence

2. 02What actually happens when I use it?

   One screenshot with real data in it, or three literal steps. Not a feature grid

3. 03Will it handle my awkward case?

   The honest edge-case section, including the cases you do not handle yet

4. 04Has anyone else done this and survived?

   Proof, placed here rather than in the hero

5. 05What does it cost, and what happens if I stop?

   Pricing, plus one line about cancellation and getting your data out

6. 06What do I do right now?

   One button, describing an outcome rather than an act of admin

Writing those section headings as questions in the reader's own words does two jobs at once. It keeps you honest about whether the section earns its place, and it makes the page legible to the machines that increasingly read it before a human does. A heading that matches the question somebody typed is the part an assistant can lift and attribute.

The section founders resist writing is the one about the awkward cases, and it is consistently the one that converts. Naming what your product does not do yet costs you the readers who were going to churn in week two, and it buys credibility with everyone else, because you have just said the one thing a marketing page is never supposed to say.

Where does proof go when you have almost none?

When you have almost no proof, make the proof slot small, specific and true rather than large, generic and borrowed. One named quote from one real beta user, with their role attached, does more work than six logos you are not entitled to display.

There is a rough ladder here, and it is worth knowing which rung you are actually standing on.

• A named person, a role, and a specific sentence. The strongest thing available to a pre-launch product, and the easiest to get, because it costs your user two minutes.
• A small number you can defend. Eleven practices for four months reads as real precisely because it is unimpressive. Round, large, unsourced numbers read as marketing.
• Your own working, shown in public. A changelog, a build log, a short demo of the real product with real data in it. Evidence of competence substitutes for evidence of demand.
• Logos and trusted-by lines you have not earned. A visitor who recognizes none of them concludes the trust was borrowed, and that suspicion then travels back up the page.

There is a newer reason to care about the exact sentence in that slot. Your positioning no longer gets read only on your page. In Similarweb's analysis of the domains ChatGPT cites most, Wikipedia and Reddit sit far ahead of everything else at 13.15 and 11.97 percent of citations, which is more than a quarter of them between two community and reference sites. Your product gets described secondhand, by other people, in places you do not control. Write the one sentence you want repeated, put it where somebody can copy it, and make it short enough to survive being pasted into a comment.

Getting that first quote is a distribution problem rather than a writing problem, and it is the one covered in how to get testimonials for a new product.

What should the button actually say?

A call to action should name one outcome the reader wants, not one act of administration you need. Get started, sign up and learn more all describe work the visitor is being asked to do. The replacement is usually the same click described from the other side of it.

One action per page, repeated. Two competing buttons of equal weight is the most common self-inflicted wound on a founder landing page, because it converts a decision the visitor was ready to make into a decision they now have to think about. Put the secondary path in text at half the visual weight, and let the primary button carry the page.

Three before-and-after teardowns

Here is the same reasoning applied end to end. Each pair below takes one page, shows the copy that was there and the copy that replaced it, then explains the three decisions in the margin. Each teardown targets a different part of the page, because founder pages fail in different places and a checklist that only fixes heroes leaves most of the problem standing.

Read the three sets of margin notes together and the same move shows up in all of them. Every fix replaced something the founder believed about their product with something the reader could verify in a second. That is most of the craft. Copy is not persuasion so much as the removal of every sentence a stranger has to take on faith.

How do you test copy with no traffic?

You cannot split-test forty visitors, and pretending otherwise wastes weeks. The substitute for a bigger sample is a different instrument: ask people directly, in a structured way, and treat five answers as the data.

Run it like this and it takes an afternoon.

• Five people, five seconds. Show the page, close it, then ask what it does, who it is for, and what they would do next. Write the answers down verbatim rather than summarizing them into what you hoped they meant.
• Two wrong answers out of five is a failing hero. Not a debate, a threshold. Rewrite the headline and the subhead, then run it again with five different people.
• Ask the right five. Friends and designers give you taste feedback. You want people who have the job your product is for, or founders who ship pages often enough to spot the pattern from cold.
• One section per pass. Rewriting the whole page at once means you never learn which change did the work, even at this small scale.

The awkward part is finding five people who will be honest with you. Machine feedback gets you the first pass for free: our landing page roast tool reads a page and tells you where the hero fails the five-second test, which is the same critique I would give and takes about thirty seconds. Why we built that as a free public tool rather than a lead magnet is laid out in the weekend free-tool playbook.

Human feedback is the part that is genuinely hard to buy, which is a large part of why Favors.dev exists. It is a founder marketing co-op with a points economy: you earn points by giving other founders real, specific feedback on their pages and launches, and you spend those points to have a room of founders pull yours apart. Nobody is being polite to you out of obligation, which is exactly why the notes are usable. Once the page is right, the next constraint is getting the right traffic in front of it, which is a sales problem rather than a copy one and is covered in your first 10 paying customers. The wider set of zero-budget channels that feeds a page like this sits in SaaS marketing on no budget, and the checklist version of a roast, including what a good one should cover, is in roast my landing page.

Frequently asked questions