Social Proof Examples, Torn Down: 12 B2B Websites

Twelve B2B software homepages, torn down one by one: Retool and Clay put a named customer and a specific dollar or percentage figure directly in the scroll, which is the pattern worth copying. Linear, Salesloft, Pipedrive and Vercel defer their proof past the point most visitors ever scroll to, or hide it behind JavaScript a search engine cannot read at all. Most B2B sites sit somewhere between the two: a logo wall up top, a number buried near the footer.

How we did this

We fetched the public homepage of each site below exactly as a search engine or an AI answer engine would (no login, no JavaScript execution), the same constraint our B2B social proof post argues actually matters: content that only renders after a script runs is invisible to the crawlers that decide what gets cited. For each site we recorded what proof appeared, roughly where on the page (as a percentage of scroll depth through the raw page content), and whether it carried a specific number or only a logo and an adjective. Quotes below are short fragments, under 15 words, with a link to the live page for the rest.

The twelve, at a glance

Five of twelve pair a specific number with a named customer anywhere on the homepage. Two return no readable proof at all in the raw page.

The twelve, in the order they impressed us least to most

Pipedrive (pipedrive.com). Nothing. No logo strip, no testimonial, no rating, no customer-story link anywhere in the page content a crawler actually receives. If Pipedrive has social proof on its homepage, it only exists after JavaScript runs, which means it does not exist for the purposes of search or an AI answer engine. Same failure mode as leaving the section out entirely.

Vercel (vercel.com). The raw page delivers almost no text at all, under 4,000 characters versus 10,000 to 18,000 for most sites in this set. Whatever proof exists is rendered client-side. The most JavaScript-dependent homepage in this set. A famously polished product with a homepage that is functionally blank to anything that does not execute a script.

Salesloft (salesloft.com). One line, "Customer Stories, See how other companies use Salesloft," a nav-style link with no name, no number, and no quote anywhere on the homepage itself. Notable because Salesloft sells sales enablement, the exact discipline that argues customer proof belongs in front of a buyer, not one click away from it.

Linear (linear.app). "Powers over 40,000 product teams, from ambitious startups to major enterprises, Customer stories," the very last content block before the footer. A real, large number with zero supporting quote or named company anywhere on the page, positioned after a reader has almost certainly stopped scrolling. Linear's actual pitch is its interface; the proof is an afterthought.

Miro (miro.com). "20,000+ reviews from Capterra, G2 and Trustradius," next to a row of compliance badges (SOC 2, ISO 27001, GDPR). A separate "Customer Stories" nav link exists but nothing from it appears on the homepage itself. Aggregated review-count proof and compliance badges are a real signal, but they answer a different question than a named customer would. Nothing on the page says what any specific team achieved.

Amplitude (amplitude.com). "Trusted by teams shaping the future," a genuinely strong logo wall (Anthropic, NVIDIA, Coinbase, Cloudflare, Mercury, Character.AI) directly under the hero. The logos alone are the pitch. Recognizable AI-native brands do real work here, but no number or named result sits next to any of them.

Ramp (ramp.com). A dense, credible logo wall (Shopify, Perplexity, Notion, Eight Sleep, Foursquare) rendered as images with no adjacent claim text a crawler can read. Same brand-recognition strategy as Amplitude, executed as pure logo, no headline sentence at all. Effective for a human scanning the page, invisible as text to anything else.

ClickUp (clickup.com). An early "Trusted by the best" logo strip, then a dense proof stack near the bottom: a named quote ("ClickUp is serving us so we can serve our pet guardians," a Senior Project Manager), "Rated 4.7/5 by 10,000+ users on G2," and "85% of Fortune 500 companies." More proof types than any other site here (logos, testimonial, third-party rating, adoption stat), but almost all of it sits past 85% scroll depth. A visitor who does not scroll to the bottom sees only the opening logo wall.

Attio (attio.com). "Trusted by 30,000+ customers," immediately followed by a specific result: "Artificial Intelligence: 83% faster lead triage, how Granola turns product signal into pipeline." The pairing (an aggregate customer count next to one named customer's specific number) is a genuinely good format. The flaw is placement: 89% down the page is nearly the same failure as Linear's, just with better content once you get there.

Webflow (webflow.com). An early, diverse logo wall (The New York Times, TED, Spotify, Dropbox, Orangetheory), then by 24% scroll depth a specific figure: "$200M in new pipeline generated post-site launch," attached to a named customer (ABM) with a "Read the story" link. The best-placed proof in this set: a recognizable logo wall early, then a quantified named result before a quarter of the page has gone by. A visitor does not have to commit to reading the whole homepage to see it.

Clay (clay.com). "Trusted by more than 500,000 leading GTM teams," then two named, quantified story teasers embedded directly in the scroll: "How Hex increased inbound win-rate 50% using Clay" and "How Verkada GTM team expanded in 28 European countries using Clay," both with a direct "Read case study" link. Two separate named customers, two separate specific results, both with their own link, not bundled into one generic "customer stories" block. The 500,000 figure is a scale claim, not a result, but the two case teasers next to it do the actual persuading.

Retool (retool.com). "Trusted by 10,000+ teams," then three separate result lines in a row: "Ramp saved $8M and 20,000+ hours," "DoorDash saved $6M and 36,000+ hours," "$3M+ profit generated and 80% faster development," each with its own "Read story" link, plus a logo wall including Amazon, Boeing and Lyft. The strongest pattern in this set. Three named, dollar-and-hour-specific results stacked together, each independently linked, backed by an enterprise-credible logo wall. A skeptical reader gets three separate chances to find a company that looks like them, not one generic claim to take on faith.

Six patterns across all twelve

A number next to a name beats a number alone, every time we saw it. Retool's "Ramp saved $8M" and Clay's "Hex increased inbound win-rate 50%" work because the reader can look up the company and decide if it resembles them. Linear's "40,000 product teams" and Clay's "500,000 GTM teams" are scale claims with no name attached, and they read as marketing copy even when the underlying number is real.

Placement matters more than most homepages act like it does. Webflow and Amplitude put real proof inside the first third of the page. Attio and ClickUp have equally good proof, sometimes better, but push it past 85% scroll depth, where most visitors never arrive. The content existing is not the same as the content being seen.

Two of twelve sites had no crawlable proof at all. Pipedrive and Vercel may well have testimonials, ratings or logos on their live pages, rendered by JavaScript after load. A visitor sees them. A search crawler or an AI answer engine, which is exactly how our own Customer Proof Index was built, does not. If your proof only exists after a script runs, it does not exist for the systems deciding whether to cite you.

Logo walls are the cheapest form of proof, and most sites still lean on them alone. Ramp, Amplitude and the opening of ClickUp and Webflow all use recognizable brand logos as their first proof signal. That works for brand recognition. It does not tell a reader what happened, which is why the sites that paired a logo wall with a specific, named result (Webflow, Clay, Retool) did more work with the same amount of homepage space.

A specific mechanism beats a specific feeling. Retool's quotes state what happened ("saved $8M and 20,000+ hours"). ClickUp's featured quote ("ClickUp is serving us so we can serve our pet guardians") is warm but leaves out what changed operationally. Both are real customer voices; only one gives a reader evidence to evaluate rather than a vibe to trust.

A third-party rating is a different kind of proof than a named case study, and the strongest homepages use both, not one instead of the other. ClickUp's "Rated 4.7/5 by 10,000+ users on G2" and Miro's "20,000+ reviews from Capterra, G2 and Trustradius" are aggregate, independently verifiable signals a single customer quote cannot provide. Neither ClickUp nor Miro relies on the rating alone: ClickUp pairs it with a named quote and an adoption stat, while Miro's homepage leans on the aggregate rating precisely because it does not surface a specific customer result anywhere else on the page, which is the weaker version of the same idea.

How to check your own homepage against this

Five questions, in the order we'd check them, using the same constraints a crawler faces rather than what you see with JavaScript enabled in your own browser.

Fetch your own homepage with curl, not your browser. If your social proof section is empty in that output the way Pipedrive's and Vercel's were, it does not exist for a search engine or an AI answer engine, regardless of what a human visitor sees.

Find the first specific, named result on the page and note its scroll position. If it is past the halfway point, you have Attio's and Linear's problem: real proof that most visitors never reach.

Check whether any claim states a number with no name, or a name with no number. Both halves are needed. A logo wall alone (Ramp, Amplitude) reads as marketing. A big adoption number alone (Linear's 40,000 teams) reads the same way.

Count how many separate named customers appear, not how many logos. Retool's three (Ramp, DoorDash, and a third account) each get their own line and their own link. One generic "trusted by our customers" block, however many logos sit under it, is one data point, not several.

If the honest answer to any of the four checks above is "we don't have that," that is the gap a single customer interview closes, not a redesign. Every example in this teardown that worked (Retool, Clay, Webflow) started from one customer willing to say a specific sentence out loud.