We Analyzed 316 SaaS Landing Pages. Here Is the Structure They Actually Use

Most landing page advice is vibes. "Add social proof early." "Keep the hero short." Useful, maybe, but rarely backed by anything you can count.

So we counted. Rifframe's section catalog is extracted from real websites, and building it required scraping and classifying a lot of pages. This article shares the raw numbers from that work: 316 landing pages from modern SaaS, dev tools and startup products, decomposed into 1,850 classified sections.

The dataset and the method

We collected 316 landing pages between late 2025 and early 2026. The list skews toward products that designers cite as references: Linear, Attio, Stripe, Vercel, Clay, Cal.com, Raycast, Supabase and a long tail of seed-stage SaaS, dev tools and creator products. 311 pages scraped cleanly; 5 failed (parking pages and coming-soon screens).

Each page was segmented into visual sections, and each section was classified by type using its structure: heading patterns, image counts, CTA counts and bounding box heights. The classifier is conservative: when it was not sure, it labeled the section unknown rather than guessing.

That gives us 1,850 classified sections, about 5.9 detected sections per page. The real average is a little higher because the conservative classifier leaves 19 percent of sections unlabeled, but the relative proportions are solid.

What a landing page is actually made of

The distribution across the 1,850 sections:

Two things jump out.

First, features sections dominate everything. More than a third of all classified sections exist to explain what the product does. The average page does not have one features section, it has two. Explaining the product is most of the job.

Second, testimonials are rarer than the advice suggests. Only 2 percent of classified sections are testimonial blocks, and many strong pages skip them entirely. The references compensate with other proof: logo strips, usage numbers, and increasingly the product itself shown in the hero. Social proof matters, but a wall of quotes is not the only way to deliver it.

The hero: three patterns rule

Within the 235 hero sections, three layouts cover the large majority of modern SaaS pages:

Centered with screenshot below

Title and subtitle centered, two CTAs, then a large product screenshot under everything. This is the dominant pattern among current B2B SaaS references: Linear, Attio and Amplemarket all use it. It works because the product is the proof, and it pushes the screenshot above the fold without fighting the copy for horizontal space.

Split text and screenshot

Copy on the left, product visual on the right, roughly 50/50. The classic B2B layout. Shorter than the centered pattern (700 to 900 pixels versus 1,100 to 1,500), which matters if you want the next section visible on a laptop.

Floating cards

Centered copy with small product cards floating around it. Almost exclusive to consumer and creator products (Buy Me a Coffee is the canonical example). If you sell to developers or businesses, you almost never see this one.

Across all three patterns, two micro-elements repeat constantly: an eyebrow badge above the title ("New: feature X", "Backed by Y Combinator") and a trust line under the CTAs ("No credit card required", "5,000+ teams"). Cheap to add, present everywhere.

Features: three formats, one rule

The 658 features sections break down into three recurring formats:

• Zigzag: text and screenshot alternating left and right, one row per feature. The workhorse of product-led pages.
• Bento grid: one large cell plus several small ones, asymmetric. The fashionable format since Stripe and Apple popularized it.
• Three-column grid: icon, title, two lines of text. The oldest pattern and still everywhere, because it scans in two seconds.

The rule that emerges from the references: dense formats are followed by light ones. A bento grid is followed by a single quote or a stats row, not by another bento. Pages that stack two dense sections in a row read as noise.

The structure that emerges

Put the numbers together and the median modern SaaS page looks like this:

1. Nav (floating, almost always)
2. Hero (centered with screenshot, eyebrow badge, trust line)
3. Logo strip or usage numbers
4. Features, format one (often zigzag)
5. Features, format two (often a grid, lighter)
6. Something that breaks the rhythm: stats band, single quote, demo video
7. Pricing or a comparison, when the product sells self-serve
8. FAQ
9. Final CTA band
10. Footer

Nine to ten sections. Not five, not twenty. And the order is remarkably stable across the dataset: attention, then explanation, then proof, then action.

Methodology notes, honestly

Automated section classification is imperfect; 19 percent of sections did not fit a confident label, and testimonial detection in particular probably undercounts quote-style blocks embedded inside other sections. The dataset also skews toward design-forward products, because that is what we wanted to extract patterns from. Treat the percentages as strong directional signals, not decimals to defend in court.

Every pattern named in this article exists as a buildable section in the Rifframe catalog, extracted from this exact dataset. If you want the nine-section structure above without assembling it by hand, the editor builds it from a one-line brief.