Sixty-two funded SaaS homepages. One scored above 90.

• The bar is Google's own: a "good" mobile LCP is 2.5 seconds. The median site in this sample took 8.45, more than three times over.
• It is not a few stragglers. 30 of 62 sat in the red zone, a score under 50, and 52 of 62 needed more than 4 seconds to show their main content.
• The tail is worse than the median suggests: 25 sites took over 10 seconds, and the slowest hero measured 82 seconds on a phone.
• Exactly one of the 62 cleared 90 and hit the 2.5-second bar. Fast is possible; it is a choice most of these teams have not made yet.

Before we send a cold email, we run the recipient's homepage through Google PageSpeed Insights on its mobile setting — the same free test anyone can run. Two numbers do most of the work. The performance score (0–100) is Google's rollup of how the page loads; anything under 50 is flagged red. Largest contentful paint (LCP) is the moment the biggest thing on screen — usually the hero headline or image — actually finishes rendering. Google's threshold for "good" is 2.5 seconds. Past 4 seconds it is graded "poor."

The sample here is 62 funded B2B SaaS startups — 30 seed, 32 Series A — measured on 2026-07-01 and 07-02. One honest caveat, stated up front: this is not a random sample. We found these companies during outreach research, and part of our filter favored heavy, marketing-led homepages that were likely to be slow. Read it as a floor check on a specific slice of the funded-SaaS web, not an industry average. The distribution still says something the averages hide.

The performance scores split almost in half. 30 of the 62 scored under 50 — Google's red zone. Another 31 sat in the 50–89 middle. Exactly one reached the green 90+ band, topping out at 97. The median came in at 50.5: a coin-flip page, right on the line between "needs work" and "poor."

Load time is where it gets concrete. The median hero took 8.45 seconds to paint on a phone. Only one site in 62 met Google's 2.5-second bar; 52 were over 4 seconds, and 25 were over 10. The worst cases were not typos in the data. An AI sales-rep platform took 82 seconds to render its hero; a recruiting-automation tool took 68. One corporate-card fintech was subtler: it painted in about 7 seconds, then stayed frozen to taps for another 34, the longest blocking time in the set. A page can finish painting and still answer nothing.

The pattern underneath is familiar: a homepage assembled in a page builder or heavy framework, stacked with autoplaying video, uncompressed hero art, and a dozen marketing scripts, each shipping a little more JavaScript to a device with a fraction of a laptop's power. None of it is visible on the founder's fast office connection. All of it is visible to the buyer on a train.

A landing page has one job: convert traffic into a demo, a trial, or a signup. That traffic is not free — it arrives from ads, from a launch, from a founder's hard-won post. Every second the hero is blank is a second the visitor is looking at nothing, deciding whether to wait. Google's own field data puts the bounce probability up sharply as mobile load climbs past a few seconds; at 8, a meaningful share of the paid click is simply gone before the pitch begins.

The cruel part is that the money is already spent. The click was paid for, the visitor arrived — and the page handed them a spinner. Speed is not a vanity metric on a landing page; it is the first conversion step, the one that happens before a single word of copy gets read. A page that loads is table stakes for a page that sells.

One page in 62 scored 97 and painted its hero in 2.5 seconds. It was not running a lighter product or a smaller company — it had simply been built with load time as a constraint rather than an afterthought: restrained JavaScript, compressed and correctly-sized images, the important content shipped first. Nothing exotic. The gap between that page and the 82-second one is not budget or talent. It is whether anyone made speed a requirement before the page shipped.

For reference, the page you are reading is hand-coded to the same discipline we hold client work to: Lighthouse 97+ performance, 100 on accessibility, best practices and SEO, under 290 KB fully loaded, no framework and no page builder. Every number on this site is measured, never estimated, and re-measured at every deploy.

More writing: the Linear teardown →