Website Speed and Conversion Statistics, With Sources

There are three genuine primary studies behind almost every website statistic you have ever been quoted. Everything else is those three, reworded.

This page names them, links them, gives the sample size, and does the thing the rewrites never do: states how old each study is, and where its data stops applying to a contractor site.

That last part matters more here than on most statistics pages. Two of the three studies measure ecommerce checkouts. You do not run a checkout. A homeowner filling in a foundation inspection request behaves differently from someone buying trainers, and pretending otherwise is how a benchmark becomes a bad decision.

Load time against conversion rate

The strongest study in this category is Portent’s site speed research. Sample: 20 websites, over 27,000 landing pages, more than 100 million page views, with load speed measured across 5.6 million sessions over 30 days. Six of the sites are B2C ecommerce. Fourteen are B2B lead generation.

Take the lead generation half. That is the closer analogue to a contractor site, because the conversion is a form fill or a call, not a purchase.

• A site loading in 1 second converts 3x better than a site loading in 5 seconds
• A site loading in 1 second converts 5x better than a site loading in 10 seconds

The ecommerce half, for completeness: conversion drops an average of 4.42% for each additional second between 0 and 5 seconds. Sites loading in one second converted at 3.05%, falling to 1.08% at five seconds.

One more finding from the same study that gets left out of every summary: 82% of the pages measured loaded in 5 seconds or less, unchanged from when Portent ran the study in 2019. The web did not get faster. If your site is slow, you are in a real minority, and working out whether it actually is takes about five minutes.

Load time against bounce rate

From Google’s mobile speed research, published through Think with Google:

• Bounce probability rises 32% as load time goes from 1 second to 3 seconds
• Bounce probability rises 123% as load time goes from 1 second to 10 seconds
• 53% of mobile visitors abandon a site that takes longer than 3 seconds
• 70% of the mobile landing pages analysed took more than 5 seconds to show above-the-fold content, and more than 7 seconds to load fully

Date this one before you use it. The underlying research is from 2016-2018, which makes it the oldest thing on this page by some distance. Phones, networks and browsers have all changed since. The 53% in particular gets quoted as current fact constantly and it is nearly a decade old.

The direction of the finding has held up in later work, including Portent’s. The specific percentages should be treated as historical.

How many sites actually pass Google’s speed thresholds

The most current and most transparent dataset on this page, and the one nobody quotes, is the HTTP Archive Web Almanac 2025 performance chapter. It combines Chrome UX Report real-user data with lab measurement across millions of websites, based on July 2025 measurements. It is free, the methodology is published, and it is rerun every year.

48% of mobile origins pass all three Core Web Vitals. 56% of desktop origins do.

The trend on mobile, which is where your homeowners are:

Year	Mobile origins passing all three
2023	36%
2024	44%
2025	48%

Broken down by metric, on mobile:

• LCP (how fast the main content appears): 62% good
• INP (how fast the page responds when tapped): 77% good
• CLS (whether the layout jumps around): 81% good

Loading is the bottleneck. Roughly four mobile sites in ten fail on LCP alone, and LCP is the metric a homeowner on a phone in a wet basement actually feels.

Two useful things fall out of this. First, more than half of mobile sites still do not pass, so failing is normal rather than shameful. Second, the bar is rising about four points a year, which means standing still is losing ground slowly.

If you want your own numbers rather than the population, Core Web Vitals for your domain are free to check in Google Search Console, and they are real user data from actual visitors rather than a lab test.

Form length

From the Baymard Institute, which runs large-scale usability testing rather than surveys:

• The average US checkout shows 23.48 form elements, or 14.88 actual form fields, to a new customer
• Most checkouts could cut form elements by 20-60% without losing anything
• Average documented cart abandonment is 70.22%, calculated across 50 separate studies
• Baymard estimates the average large ecommerce site could gain 35.26% in conversion through better checkout design

Baymard’s own observation from testing is the useful part: users are consistently overwhelmed and intimidated by a high number of form fields and selections.

Now the caveat, and it is a big one. Every figure above measures an ecommerce checkout. A contractor quote form is not a checkout. Nobody abandons a basement waterproofing enquiry because shipping was too expensive, and the 70.22% abandonment rate does not transfer to your estimate request in any meaningful way.

What does transfer is the finding underneath the numbers: field count suppresses completion, and most forms carry fields nobody needs. That principle is why the fields on a contractor quote form are worth arguing about. The percentages are not yours to borrow.

Where the traffic comes from before it hits your site

One figure from the BrightLocal Local Consumer Review Survey 2026 (1,002 US adults, published 11 February 2026) belongs on this page rather than with the review statistics:

54% of consumers visit the business website after reading positive reviews. It is the single most common next action, ahead of reading more reviews (37%) or contacting the business directly (20%).

Reviews and rankings do not book jobs. They hand a warm homeowner to your website. Whatever happens next is a website problem, which is the whole reason local SEO and site quality cannot sensibly be bought as separate projects.

What is missing from all of this

There is no published, credible, large-sample study of conversion rates on contractor websites specifically. We looked. What circulates instead is agency blog posts quoting an average conversion rate for “home services” with no sample, no method and no date.

If you see a claim that contractor websites convert at 3.2%, or 5%, or any tidy figure, ask where it came from. In every case we chased, the answer was another agency blog.

This means the honest position on your own site is that you have no benchmark. You have a before and an after. Measure calls and form fills per hundred visitors for a month, change one thing, measure again. That is a worse story than a benchmark and a far better basis for a decision.

Where these numbers do not help you

Speed is a ceiling, not an engine.

A one-second load time on a site that does not say which towns you serve, does not show a phone number on mobile, and has four photos of a truck will still not book work. Portent’s 3x is the difference between a fast site and a slow site that are otherwise the same. It is not a promise that shaving two seconds triples your calls.

The order that usually holds for a contractor: the site has to say the right things, then it has to be fast, then it has to be easy to convert on. Sites get rebuilt for speed all the time when the actual problem is that the site never asks for the call.

And if your traffic is in the low hundreds a month, none of these percentages will be visible in your numbers for a long time. At that volume the constraint is almost never the site. Getting a high-converting website for contractors built before there is anything to convert is a common and expensive order of operations.

Using these statistics

Portent, Google and Baymard all publish their research free. Cite them directly rather than this page.

What this page adds is the age of each study and the point where its data stops applying to a contractor site, which is the context those figures usually travel without.