Ten SaaS homepages, five that convert and five that don't, with what makes the difference.

The setup

Median SaaS landing page conversion is 3.8%. Top-quartile is 11.45%+. The companies below mostly sit at one end or the other. The ones converting at the top end share four traits: clear positioning, product visible in the hero, restrained CTAs, named social proof immediately. The ones in the low-converting group fail at least two of those, usually three.

Worth naming the disclaimer upfront: the second list isn't picking on bad companies. They're real businesses with real products and real customers. The point is to make the patterns concrete so they're easier to spot on your own site.

The five that convert

1. Stripe (stripe.com)

What it does well: Stripe's homepage is the platonic ideal of a B2B SaaS homepage in 2026. The italicized hero ("Financial infrastructure to grow your revenue.") anchors a subhead that's specific in every direction: "Accept payments, offer financial services, and implement custom revenue models, from your first transaction to your billionth."

That subhead is doing five jobs at once. It names three capabilities (payments, financial services, custom revenue models), frames the audience (any business with online revenue), signals scale (first transaction to billionth), suggests both small-startup and enterprise fit, and uses concrete numbers instead of adjectives.

Directly under the hero: a scrolling logo marquee with Amazon, NVIDIA, Ford, Google, Shopify, Figma, Uber, Anthropic, OpenAI. Then a stats band: "135+ currencies, $1.9T processed in 2025, 99.999% uptime, 200M+ subscriptions."

The signature move: italics as the entire emphasis system. No bold. No color tricks. Just italic. It's the typographic equivalent of confidence. The case study photos compose architectural elements into Stripe's parallelogram logo, a level of design discipline almost nobody else attempts.

The pattern to copy: specific subhead doing five jobs at once, social proof immediately under the hero, one tight typographic emphasis system, no decorative complexity.

2. Linear (linear.app)

What it does well: Linear's homepage is a working theatrical demo. Not screenshots, a fake-but-functional issue (ENG-2703 "Faster app launch") with Slack threads, Codex agent terminal output, and an embedded MP3 "Weekly Pulse" podcast you can actually play.

The hero is short: "The product development system for teams and agents." Subhead: "Purpose-built for planning and building products. Designed for the AI era."

The audience signal is precise. Anyone reading "product development system" who isn't in product development self-selects out. The ones who stay are exactly the buyers Linear wants. Names dropped throughout: OpenAI's Gabriel Peal, Ramp's Nik Koblov, Opendoor's Kaz Nejatian.

Dark mode default with status colors borrowed from the app itself (yellow Todo, blue In Progress, purple In Review, green Done). Engineering-journal section labels ("FIG 0.2"-style). Inter Variable at huge sizes with negative tracking.

The pattern to copy: the homepage IS the product. Not screenshots, working surfaces. The most-imitated SaaS site of 2024–2026 for good reason.

3. Loom (loom.com)

What it does well: The CTA is just play-the-video. An embedded autoplaying Loom that is the demo. The product demonstrates the product by being the product.

Headline: "One video is worth a thousand words." (Grade 4 reading level, fewer than 7 words.) Subhead is tight. The video does the rest.

Below the hero: "Millions of people across 400,000 companies choose Loom" + a 25+ logo wall (HubSpot, Lacoste, LaunchDarkly, Atlassian, Brex, Ford, Mercedes-Benz, Tesla, Volvo, Goldman Sachs, Asana, Walmart, Disney). The logos themselves do positioning work, if Goldman Sachs and Disney both use it, the product is enterprise-credible AND consumer-friendly.

The isometric illustrated portraits for testimonial avatars (instead of headshots) is a signature visual choice that's survived the Atlassian acquisition.

The pattern to copy: if your product is visually demonstrable, the demo IS the hero. Don't separate them.

4. Vercel (vercel.com)

What it does well: Total grayscale discipline. Color is reserved for product output, not chrome. The page renders live AI Gateway telemetry in the hero, "Top models on Jun 2/3, 2026: Gemini 3 Flash, DeepSeek V4, Claude Opus 4.8", pulled from real customer usage. Product data doubling as marketing.

The hero is short: "Build and deploy on the AI Cloud." Subhead: "Vercel provides the developer tools and cloud infrastructure to build, scale, and secure a faster, more personalized web."

No traditional logo wall. Instead, a single rotating sentence inline: "Runway build times went from 7m to 40s. LeonardoAi saw a 95% reduction in page load times. Zapier saw 24x faster builds." Three sentences. Three named customers. Three specific metrics. More powerful than 20 logos.

The pattern to copy: the inline single-sentence social proof format. One specific number from a recognizable customer beats a logo bar for converting technical buyers.

5. Cursor (cursor.com)

What it does well: Named pull-quote testimonials from operators the visitor would actually recognize. Jensen Huang (NVIDIA, "some 40,000 engineers"), Patrick Collison (Stripe, "thousands of extremely enthusiastic Stripe employees"), Andrej Karpathy, Greg Brockman (OpenAI), shadcn.

Hero: "Built to make you extraordinarily productive, Cursor is the best coding agent." One sentence. No subhead acrobatics.

Most surprising design choice: commissioned oil-painted landscape wallpapers behind product mocks. Utterly off-script for a developer tool. It works because everything else is so disciplined that the painted backgrounds become signature instead of distracting.

Dark mode with warm umber/charcoal (not pure black). Custom geometric sans (Cursor Sans, built by Kimera). Footer: "Trusted by over half of the Fortune 500."

The pattern to copy: named testimonials from people the visitor recognizes outrank logo bars for converting senior buyers. A single named quote from Jensen Huang does what 50 logos don't.

What the five share

Sharp positioning, product visible in the hero, social proof immediately, one primary CTA, specific subhead doing 4–5 jobs at once. None of them are accidents. Every detail is a deliberate choice. The patterns repeat across all five because they actually work, not because the designers copied each other.

The five that don't convert as well as they should

A different list. These are real, working SaaS companies, I'm not picking on small ones. Each demonstrates a specific pattern that suppresses conversion. (Light anonymization where the criticism gets specific, since the point is the pattern, not the company.)

1. The mid-market enterprise software site

The pattern: feature-led hero, no product visible.

Hero says some variation of "Unified enterprise [category] platform" with a marketing render that's not the actual product. Subhead lists three capabilities. CTA is "Talk to sales" with no starting price.

What's wrong: nothing in the first viewport tells the visitor what the product looks like, who it's for, or what it does that competitors don't. The page assumes the visitor came in already qualified by sales. For warm-intro traffic, that works. For cold traffic, it converts at 0.5–1%.

The fix: Add a real screenshot. Name a category. Put pricing somewhere on the page, even if it's just "starting at."

2. The well-funded AI startup that hasn't shipped marketing

The pattern: hero is a single buzzword sentence with no audience or outcome.

"AI for the future of work." or "Intelligence at the speed of business." Aesthetic gradient. No product visible. One CTA: "Join the waitlist."

What's wrong: visitors can't tell if this is a B2B SaaS, an API for developers, a consumer app, or a research lab. The waitlist signups will be tire-kickers, not qualified prospects.

The fix: name a category. Name an audience. "AI customer support for B2B SaaS teams" converts 10x better than "AI for the future of work," even if the underlying product is identical.

3. The consulting-firm-shaped SaaS site

The pattern: copy that reads like a McKinsey deck.

"Leveraging proprietary machine learning algorithms, our enterprise-grade platform empowers organizations to optimize workflow efficiency through unified observability and intelligent automation."

Hemingway scores that at grade 14. Unbounce's data is clear: copy at college reading level converts at half the rate of copy at 5th–7th grade level. The same product, described in simpler language, would convert measurably better.

The fix: strip every adjective. Use verbs. Aim for grade 6. Read the result out loud, if it sounds like a textbook, rewrite.

4. The agency-sold SaaS site

The pattern: 5+ CTAs in the first viewport.

"Start free trial" + "Book a demo" + "Watch product video" + "Talk to sales" + "Read the whitepaper." All weighted similarly. The visitor gets decision fatigue and bounces.

What's wrong: when an agency builds a SaaS site for a CRO who wants every funnel covered, every funnel gets a CTA on the homepage. The result converts worse than any single one of those funnels would alone.

The fix: pick one. The single highest-value action you want a visitor to take. Demote everything else to text links or moves it to dedicated pages.

5. The "we're for everyone" site

The pattern: trying to convert five different ICPs on the same homepage.

Hero: "A unified platform for marketing, sales, engineering, finance, and operations teams."

What's wrong: nobody recognizes themselves on the page because the page is for everyone. April Dunford's litmus: "If you removed your logo from your homepage, would people still recognize it as your product?" For a "we're for everyone" site, the answer is no, because the page would also describe HubSpot, Notion, Monday, ClickUp, and 20 other tools.

The fix: pick the single best-fit ICP. Write the homepage for them. Build separate landing pages or sections for the other audiences. The homepage gets to be focused; the funnel gets to be specific.

The cross-cutting patterns

The five that convert share four traits the five that don't all lack:

1. One identifiable ICP in the hero, not "modern teams."
2. Product visible in the first viewport, real screenshot, video, or working demo.
3. Social proof immediately under the hero, not 6 sections down.
4. One primary CTA, optionally one secondary, never more.

That's almost the entire delta. Six other patterns matter at the margin (typography, animation, pricing visibility, mobile load) but the four above are what separate the 11% converters from the 2% converters.

The single most-imitated move of 2025–2026

Linear's working-product-demo-as-hero. Notion, Cursor, Pitch, and Vercel have all adopted variations of it. Every premium SaaS site shipping in 2025–2026 features either a working surface in the hero or an autoplaying product video.

The pattern: don't show a screenshot of your product. Show your product running. Pull real telemetry (Vercel). Embed working state (Linear). Auto-play a real workflow (Loom). Drop in a fake-but-functional preview (Notion). The visitor doesn't have to imagine what it's like to use the product, they're already using a version of it.

This is the single biggest design lift available right now. If your hero has a static product render, you're behind. The shift to interactive product surfaces is the 2026 equivalent of what "above the fold" was in 2010.

What to do with this

Open your homepage. Score it against the four traits the five high-converters share:

• One identifiable ICP in the hero?
• Product visible in the first viewport?
• Social proof immediately under the hero?
• One primary CTA, with at most one secondary?

If you fail two or more, that's your starting point for the rebuild. Not the colors. Not the fonts. Not the animation. These four traits.