Key Takeaways
- Google and Deloitte measured a 21.6% increase in users reaching the form submission page on lead-generation sites from a 0.1 second speed improvement — the single most relevant conversion statistic for contractors, and almost nobody cites it.
- Bounce probability rises 32% as load time goes from one second to three, and 123% by ten seconds.
- Google's 'good' thresholds are Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, and Cumulative Layout Shift under 0.1, assessed at the 75th percentile of real page loads.
- Home & Home Improvement converts at 8.05% on paid search across 13,474 campaigns — a useful floor, since a landing page below it is losing to the average.
- Speed is not a ranking tactic; it is a conversion tactic that happens to be measured by search engines.
Almost every conversion statistic in circulation was measured on e-commerce checkouts, then reprinted as though it described a contractor's quote form. The studies below are the ones that either measured lead generation directly or set the thresholds your site is actually judged against. Where a figure comes from retail, it says so.
Does site speed actually change lead volume?
Yes, and the effect on lead forms is larger than on retail carts. Google and Deloitte's study found that a 0.1 second improvement produced a 21.6% increase in users progressing to the form submission page on lead-generation sites, alongside a 5.5% increase from homepage to the first form step and 7% more page views.
That 21.6% is the most contractor-relevant conversion statistic in the published literature, and it is almost never quoted — the retail figure of 8.4% gets recycled instead, because retail is what most marketing blogs write about.
The study covered 37 European and American brand sites and more than 30 million user sessions over 30 days, with no redesigns running during the measurement window. The 0.1 second improvement had to hold across four metrics simultaneously — first meaningful paint, estimated input latency, observed load and maximum server latency — so it is a genuine speed gain rather than a single lucky metric.
How much does a slow page cost in lost visitors?
Google's research with SOASTA put the probability of a bounce 32% higher when load time went from one second to three. By five seconds it was 90% higher, by six seconds 106%, and by ten seconds 123%. The damage is not linear and the steepest part of the curve sits in the range most contractor sites occupy.
What speed is fast enough in 2026?
Google's published Core Web Vitals thresholds are Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, and Cumulative Layout Shift under 0.1. Critically, these are assessed at the 75th percentile of real page loads segmented by device — so your average is irrelevant and your slowest quarter of visits decides the verdict.
Testing your site once on office wifi tells you almost nothing. The threshold is a 75th-percentile field measurement across real visitors, most of whom reach you on a phone, on mobile data, from a Facebook ad.
What conversion rate should a contractor site achieve?
Home & Home Improvement converted at 8.05% on paid search across WordStream's 2026 dataset of 13,474 US campaigns, at a $90.92 cost per lead and a $8.33 cost per click. That figure is a floor rather than a target: it describes traffic already searching for the service, which is the easiest traffic anyone will ever send you.
Why is paid social traffic harder to convert than search?
Because intent is manufactured rather than declared. Someone searching 'emergency plumber near me' has already decided; someone scrolling past your ad has not. That is why Meta leads cost roughly half what search leads cost and convert at materially lower rates — the cheaper lead is cheaper for a reason, and the difference belongs in your cost-per-booked-job maths.
Does the channel change what the landing page needs to do?
Entirely. A search landing page can assume the problem is understood and lead with proof and price. A paid social landing page has to establish the problem, the offer and the credibility before it asks for anything, which is why running one page for both channels almost always underperforms two.
What happens to conversion after the form is submitted?
Most of the loss happens there, not on the page. SearchLight's benchmark across 888 contractors recorded a 43.9% blended book rate on Local Services Ads — meaning even on a high-intent channel, more than half of leads never became customers. A landing page optimised into a funnel with no follow-up simply produces more unanswered enquiries.
Fix what these numbers point at
- Landing Page Audit
Ten-point conversion checklist with a grade and prioritised fixes.
- Landing Page vs Lead Form
Which one to run when traffic comes from paid social.
- Speed to Lead Calculator
What happens after the form is submitted, and what the delay costs.
- Contractor Website AI Search Checklist
The other half of a modern site — being quotable by answer engines.
- Funnel & lead management
Where the 43.9% book rate gets fixed.
Which of these numbers transfer to your business?
The Core Web Vitals thresholds transfer exactly, because they are Google's published standard rather than a study. The lead-generation speed findings transfer well, because lead generation is what contractors do. The retail figures do not transfer at all, and the benchmark conversion rates are population averages that say nothing about your specific offer.
Sources
The public studies behind the figures on this page, with what each one was used for. Last verified 2026-07-30.
- 1Google, Deloitte and 55 — Milliseconds Make Millions
Published 2020; data collected over 30 days at the end of 2019 · 37 leading European and American brand sites across retail, travel, luxury and lead generation; over 30 million user sessions; analytics setups audited and no UX redesigns during the measurement window; effects measured after a 0.1 second improvement across first meaningful paint, estimated input latency, observed load and maximum server latency. Source for every figure in the vertical table, including the 21.6% increase in progression to the form submission page and the 5.5% homepage-to-form-step increase on lead-generation sites, and the retail (8.4%, 9.2%), travel (10.1%) and luxury (40.1%) comparisons.
- 2Google (with SOASTA) — Find Out How You Stack Up to New Industry Benchmarks for Mobile Page Speed
2017 · Machine-learning analysis of mobile session data modelling bounce probability against page load time. Source for the bounce probability table: +32% from one to three seconds, +90% to five seconds, +106% to six seconds and +123% to ten seconds.
- 3Google (web.dev / Chrome) — Web Vitals
Current guidance as at July 2026 · Google's published thresholds, assessed at the 75th percentile of page loads segmented across mobile and desktop using field data. Source for the Core Web Vitals thresholds quoted throughout: Largest Contentful Paint under 2.5 seconds, Interaction to Next Paint under 200 milliseconds, Cumulative Layout Shift under 0.1.
- 4WordStream / LocaliQ — Google Ads Benchmarks 2026
April 2025 – March 2026 · 13,474 US search advertising campaigns; minimum 52 unique active campaigns per subcategory. Source for the Home & Home Improvement paid-search figures — 8.05% conversion rate, $90.92 cost per lead, $8.33 cost per click, 6.47% click-through rate — and the $66.69 all-industry cost per lead.
- 5WordStream / LocaliQ — Facebook Ads Benchmarks: New Data by Industry
April 2024 – June 2025 · 726 US lead-objective campaigns; reported figures are medians rather than means. Source for the $41.26 Home & Home Improvement cost per lead on Meta and the $27.66 all-industry median used in the channel comparison.
- 6SearchLight Digital — Home Services LSA Benchmark — Google Local Service Ads Cost Per Lead by Trade
February 2026 · 888 contractors, $6.72M tracked Local Services Ads spend, 126,650 leads, $52.7M closed revenue. Source for the 43.9% blended book rate used to show how much conversion loss occurs after the form is submitted rather than on the page.